The Extrusion Myth We All Believed
HTML-код
- Опубликовано: 27 июл 2024
- In some recent testing of the E3d Revo we observed some ripples in the force data and calculated that these must have come from the interaction of the two main gears on the dual drive extruder. Now we need to find out if this is just a poorly made clone, or if the original Bondtech design for dual drive extrusion is flawed.
3Blue1Brown Fourier Transform: • But what is the Fourie...
Patreon: / vector3d
Twitter: @Adam_V3D
Join the Vector 3D Discord: / discord
Vector 3D Shop: vector3d.shop
Tools:
Wera Hex Keys: amzn.to/3gIYdNV
Wera Ratchet and Socket Set:https: amzn.to/3gH0OrY
Knipex Cutter: amzn.to/2Wylf2M
Ifixit Tool kit: amzn.to/3BdJXES
Wire Stripper Tool: amzn.to/3kC7hoW
3M Applicator Tool: amzn.to/3jqc8ud
0:00 Extrusion Myth We All Believed
0:43 Extrusion History and Start of Dual Drive
1:25 Test Method
2:00 Fourier Analysis
2:56 Modified Clone Extruder Results
4:23 E3D Titan Extruder Results
7:32 Bondtech BMG-M Results
9:39 Understanding Results
11:32 Does Dual Drive Force More? 
Наука
![[Устарело] Гайд по доработке Ender-3/CR-10/Aquila](/img/1.gif)
I would be VERY interested to see this test with Prusa's upcoming "zero-backlash cycloidal gearbox with a large no-slip drive gear" on their new XL printer.
Agreed. testing the effect of larger gear drives would be interesting.
that's some long name XD
they changed to a planetary gearbox with 1:10 reduction
@@simwil524 Goes in line with how stepper motors work. They're the most efficient (most power per step) when they run at full speed, IIRC. Gearing that down instead of slowing it down should provide more force.
@@simwil524 Prusa begins to look more like Elon musk. He makes some ambitious claim in an attempt to innovate which makes all his fans cheer and get excited about but that does not make that much sense from a technical point of view, only to change things at the last moment when physics, budget and manufacturing constraints become a reality.
Before even seeing the video to the end, I have my view on the matter - there's a limit to how much material that can pass through the opening in the nozzle. provided, that the material is at the intended temperature. If the extruder is not able to push material through at the wanted rate it could very well be because the hotend cannot melt the material fast enough - more force will not alleviate that.
I believe you are on the right track. The OP didn't test, monitor, and assure his machine'a the hot end maintained an accurate temperature. If temperature varies, and it will, at varying plastic flow rates, the flows, pressure, and amounts are not going to be consistent.
@@BStottYeah I thinj the whole idea of using an hotend in a force test setup is flawed. I would have preferred a setup that directly measured the force
There may be a correlation between the oscillations in extrusion force and the PID Regulator of the Heating element. The high flow would cool the nozle, the temprature sinks. When the heater compensates it MAYBE hits a heat transfer limit and switches off to avoid internal overheating. Cycle Repeats.
To test this tuning of the Heating Regulation for these flowrates might prove usefull.
Nice theory but you're describing bang-bang, which is not at all PID control. A properly tuned PID controller will have very little variation in temperature. Additionally, he is using the Revo with PTC and can respond very quickly to temperature differences.
He could heat the hotend probably 20 degrees hotter and turn it off completely for the test. If the resonance frequency is gone it could be correlated to heating.
@@brisance That would be interesting to test as well. All of the PID tests I know of for 3D printers are done while not extruding, so the energy loss that is being tuned is basically just atmospheric. If you include the energy of the system lost to melting the filament (how it operates in real life), I wonder how much that could change the PID values? I’d love to see that.
In a way, I think we’re kind of slapping a bandaid on this problem (and a whole lot more) by using linear/pressure advance. That helps tune out a lot of issues, but it only changes the extruder values rather than the temperature (though a temperature response is likely way too slow to accurately mitigate the wonkyness that rheology is based on)
@@brisance Maybe according to your Systems and Controls class, but if you've ever done actual engineering work involving such systems you'll know it's quite common to just use a PID control loop to govern duty cycle for an output device that doesn't directly accept analog control (such as the power MOSFETs one uses to switch heaters -- "half" on = dead MOSFET).
@@AJMansfield1 do you not know that PID control can be implemented via pulse width modulation?
I think the overlooked factor is 1.8 degree/step motors versus 0.9 degree/step motors. And 12v vs 24v. Also don't forget the motor impedance, this causes a motor to have different torque at different speeds, and a lower or higher max speed with adequate torque.
I didn't understand everything but the conclusions you drew are very interesting and pretty clear.
The most interesting part is the conclusion that when we hear the extruder gear slipping it is probably because the required force has risen due to the inability of the hot end to melt all the material that is being inserted. In other words we are trying to put too much material in the hot end and it cannot be heated and extruded as fast as desired. Therefore the resistance generated by the hot end increases causing the gears to slip. Learned something new today!
In my case the single gear Creality brass extruder wore out really quickly and started slipping & causing lots of under-extrusion (walls not sticking together etc)
The cheap brass gear is flat & not concave to match the filament's round profile so only a very small part of the teeth makes contact with the filament on one side. A hard spring is used in an effort to compensate but this put a lot of pressure in a very small area on the teeth and wore them out quickly. I switched to a dual gear and it's way better the design just makes sense
This has inspired me to make my own test rig but with a couple additions:
Adding a high resolution rotary encoder above the extruder to measure actual filament travel, and one encoder on the extruder motor to measure its true rotation.
I'd then have 4 different dimensions (pressure, filament travel, stepper rotation, time) and could chart any 2 of them against each other. I would personally like to see measured filament speed and pressure per degree of rotation to gauge the level of extruder engagement
Could you please share your results
@@wizardOfRobots I ordered last month but still waiting on parts 😓
Do it!
PLEASE expand this into a series with more extensive testing. I would happily donate towards the idea of testing new extruders as they come out. There is a lot happening in this area and some of the issues many of us knew instinctively but couldn't really test are finally coming out.
Some obvious extruders to compare are the new bondtech LGX style gears and dual idler gear designs like the OMG.
I'd also like to see a belt extruder with no teeth touching the filament, like the concept Proper Printed is working on and hobbed drive gears that instead use opposing helical teeth (or ideally dual "herringbone" teeth) which I have not seen in the wild yet.
maybe i am mistaken, but what you wrote there looks like a huge hint at potential solutions....i dont know who else is going after this, but i will....! THANKS!!!
@@simonschneider5913 lol no worries. I've since played around with some belt-style concepts and there's definitely potential but still hurdles to overcome to keep it compact without sacrificing longevity. I'd love to see an attempt at herringbone gears. I don't have the means to attempt that but perhaps there's an existing product about the right size that could be modified....
Best of luck, I'd love to hear what you come up with if you're willing to compare notes.
This is the video I have been waiting for. Everyone has a theory on this but no data. I've found the bigger bondtech gears have helped I still can see some vertical lines but you need to look very closely.
Mihai designs does a video about the artifacts created by dual drive extruders, what causes it, and even demonstrated it.
Proper Printing has created an extruder that uses belts on either side of the filament. Would love to see you test it!
Sounds likely.
Heavy, complex, expensive. It has many disadvantages, unless you want to print with licorice.
@@sierraecho884 Hey, that's a genuine use case ,with filaments like ninjaflex.
Great, more things to beat my head against a wall ! 😃
Several things:
18mm3 is the upper limit of the hot end so the results are pretty much useles as far as extruder is concerned.
If the force is changing by changing extruder, it's weird, it should not.
Repeating this with a Volcano might shed some light, allthough i would use a 50 or 60W heater for it. As someone here mentioned, at higher flow rates even PID or weak heater will mess up results.
The 14mm3 and lower results look OK.
Thank you for the time and effort.
P.S. I print a lot of PETG and the difference in quality of parts using a normal hotend with 40W and a Volcano with 60W is huge, Volcano parts look like tasty candy.
The extruder is the only thing influencing the force if you try to squeeze more filament through the hotend than it is capable of handling. Not one of the extruders was able to feed 18mm/s through the hotend, but the dual-drive hotends were able to provide a much higher force before the filament started slipping. The big jumps in the force are the points when the filament slipped and are not influenced by the the drive train but mostly by how easy or hard the filament is able to be compressed and the lenght between the extruder and the nozzle. Very hard and rigid filament produces a high frequency and soft filament a low one. A longer path between extruder and nozzle produces a lower frequency than a long path. The test shows that the dual drive extruders are far better than the cheap one, because they "fail" at much higher resistence of the hotend and must have pushed more filament through the hotend than the single-drive one. The spikes in force are not comming from the extruders but from the filament jumping back and at that point this doesn't matter because you will have underextrusion and your hotend is your bottleneck.
@@executeOrder69 Nicely explained, thank you.
@@toma.cnc1 thank you! 😃
@@executeOrder69 I agree. Your point proves that this test isn't suited for the conclusion that was made. Either the flow rate between extruders wasn't equal or one extruder was skipping steps. Else the pushing force would've been (and must be) equal at the same flow rate.
Thanks for an awesome video explaining the relationship between everything involved in extruding some filament. It really puts things into a proper perspective for me.
To be more assured of the true flow rates it could be rather easy to weight the extruded Material given the extrusion for the same duration. The high flow rate curves still raise suspicion in me. I really like your test and presentation.
Those extruders are for very flexible, brittle, or hard to grip filaments such as Ninja flex, wood, metal filaments. They push down on both sides helping feed it straight and provide the necessary force to extrude properly. Maybe test it more as intended with a flexible filament? I know for me I couldn't print with my single and now I got a dual and prints great.
This is incorrect. You only need to apply a force on the opposite end but not grip it. It the case of flexible materials it´s not even possible since they flex and will simply stratch and deform.
Dual drive isn't claiming more PUSHING force but more GRIPPING force meaning it's less likely to grind the filament. If the filament isn't slipping then no difference in pushing force should be observed.
That second bit you wrote is kinda my point. There is no slipping, so there is no difference. The hotend reaches its limit far before the extruder so having any additional grip is basically never utilised. Titan achieved 6.3kg though with single drive extrusion. What hotend is still providing good flow output with that kind of force?
@@Vector3DP You should also consider (you likely have) that some filaments are much harder to grip. I've found that Nylon is much more slippery and needs a much tighter grip.
@@Vector3DP that extra grip does make a difference with softer and lower TG filaments to keep it from slipping or grinding. Also harder filaments where you need to dig in harder. Teeth size is a big deal too when it comes to avoid grinding the filament.
Pushing force will be down to gear reduction and stepper motor on the extruder aside from that, to state the obvious.
So to sum it up, dual gear and big drive gears is good for reliability and compatibility. Many extruders may work really good with optimized conditions. What sets them apart is how they behave in edge cases.
@@olorf Agree!
I still remember the craters my old wade extruder sometimes dug in filaments.
Now it would be interesting to test the OMG extruder (it's baffling that only My Tech Fun reviewd it) and the LGX standard or lite.
This is some great research, finding that the second gear does in fact very little is kinda eye-opening.
Though I'd love to see this test repeated with different polymers. The hardness of PLA means that any gear that digs into the filament at least a bit will require a massive force to shear the material and actually slip, but the only situations where you'll encounter those forces are during a clog or when you're going massively beyond the capabilities of the hotend. For more flexible materials, however, I could imagine a vastly different result, so having this test repeated on TPU / ABS / Nylon could unearth some more insights...
The 2nd gear makes the consistent print result worse in fact. It has to if you think about it. It is physically not possible to mesh the gears perfectly to eatch other while also to grab the filament optimally and to be adjustable. The oroginal designer probably though ah it´s not a big issue, the force is increased that´s good. But he was wrong though ....
I had really big problems with white pla, it always grinded away the filament and then did not extrude.
Now I have 0 problems with that. Also dual extruder has gears at the top and bottom so the force is applied equal and not only at the bottom. If that changes anything I don't know but I'm really happy with that.
And even when I get skipping because of too much nozzle pressure it does not eat the filament therefor I don't have any extrusion problems
Edit: the product is called micro Swiss bowden dual extruder. If u wanna check put the design.
Hey Adam,
Thanks for the work you are putting into these tests for our community. Much appreciated!
The high flow rate oscillations do not look at all like they are gear induced but rather are caused by the non-linear behaviour of the system concerning mostly filament slip, filament heat transfer and resulting viscosity based on the material melting properties. Stefan has also reported on the oscillating viscosity and flow on filament with a hotend driven to the edge of its melting capacity at several occasions. It is always an oscillating process.
When you bring the system to the edge of its capabilities it becomes highly non-linear and that make it unstable. The filament oscillates between getting partly “stuck” (high viscosity) and therefore slipping more on the gears and then heating up and flowing faster again with less force. It is a bit akin to stick-slip friction between surfaces.
If it was the motor or gears causing this then you would see the same frequencies dominate at lower flow rates with a more or less linear increase of the effect with flow rate. That is not the case. The system is stable up until a certain rate at which it suddenly becomes very prone to oscillate. You should test at only slightly lower flow rates to see the transition region.
To further test this theory you would have to try different hotends, where you would expect different breakdown flow rates and slightly shifted frequencies between them and even more shifts in frequency when you use different materials. Also with a single hotend I would expect to see some, but probably much smaller frequency shifts (peaks over distance - not time) within the transition range between stable and unstable behaviour.
Cheers,
Rob
I think what you're looking for is probably turbulent flow, not stick/slip friction.
There's no mechanism for what you're suggesting to happen, with the nozzle jamming and unjamming.
_"If it was the motor or gears causing this then you would see the same frequencies dominate at lower flow rates with a more or less linear increase of the effect with flow rate. "_
I'm not convinced that is true. If he is approaching the maximum torque the stepper can apply then it would be reasonable to expect good behaviour below some critical force, and then have the stepper miss the occasional step as you pass this threshold.
A few questions I have for this series.
1) were the e-steps adjusted between all the extruders?
2) Was it the same motor used on all extruders?
3) when running the 18 mm^3/s runs what was the mass of extrusion collected vs expected?
4) Could this be an artifact of your pwm cycle time on the hot end?
5) How would an E3D v6 with CHT nozzle compare?
Thanks for the awesome work!
My thoughts exactly.
I am guessing it is your spool holder that adds the frequency. Every time it jumps a bit further you get a ripple
I'd be very interested to see you test out the OMG extruder. Dual drive, but both drive sets are like Titans (one idler + 1 toothed)
Exactly what came to my mind when seeing the 'single drive gear' results. Would be really interested in seeing the results for it aswell.
Yeah same here, I have that on my printer and honestly like it more then bondtech. Its mainly bc mine prints nylon enclosed so the all metal holds up better with the chamber temps
Additionally the new Haldis 3d (creator of the red lizard series of hotends) extruder. similar to the OMG as a dual drive it is a dual drive AND dual gear. Would be interesting how it would compare. Unfortunately it is quite an expensive extruder (as it's designed to mainly be a hotend extruder DIrect drive combo, but it can be used Bowden style)
i would be interested too, im running the OMG on my setup and it works well, but i don't have any comparisons because i switched from the stock ender 3 pro extruder directly to the OMG ( as a direct drive setup )
For years I'd been a die hard traditional single-gear extruder bowden setup kind of guy. I wanted to do more with flexible materials so I changed one of my printers over to a Hemera. The most unexpected issue I've encountered with the hemera? Poor flow consistency - you can see ripples show up in the outer skin. This appears to be caused by the pitch and mesh of the dual-gear system in that extruder.
Looks like there may be alternate options for better quality I should look into based on what you are seeing here.
If you stay on the Bowden Setup but with a dual gear extruder, the Bowden acts as a spring and should smoothen out the extrusion. So you see these artefacts because you don't have a Bowden tube in between. So it's a combination of both.
In my experience results like that have nothing to do with gears and are fixed by a silent stepper drover upgrade. Loud stepper drivers just arent consistent enough.
@@pmcquay1 Noise generated from the driver does not affect your print quality whatsoever.
The way they step is slightly different, but the noise is not the difference in performance
@@benjaminshields9421 I changed from stock ender 3 to a silent board and the artifacts went away. Maybe the newer drivers deal with cogging better? I don't know, but myself and several other people have seen the same thing. I changed nothing else at the same time.
Also, the gears on the ms dd have something like 18 teeth, so one tooth per 5% of a turn, 5% of one turn in my klipper config right now is 1.2mm, 1.2 mm of 1.75mm filament contains 2.88 cubic mm of plastic, which would make 32mm of .2mm by .45mm print line. Suffice to say 32mm is way larger than my artifacts. My stepper motors however have 200 "teeth" per revolution, which would make for much finer artifacts.
Great work, well done data analysis. I won't repeat the other commentators, there are some points right. Just more to investigate 😄
Looking forward to your journey!
ALWAYS more to investigate with this one. I've really hit on a bottomless pit of testing this year.
I appreciate how clearly displayed you made the credits for the 3b1b clip
Try to plot the tempetature at the hot end together with the other graphs, it may be getting some time to melt, extruding the melted one, cool down a little bit, getting some time to melt...
Hey Adam, Amazing video!
I would love to see a similar test with 3D printed extruders, like Clockwork 1 and 2 and, Galileo, Sherpa etc!
Very interesting! Lots to think about here. It would be interesting to repeat the tests with a flexible filament where the drive force limits before grinding are much lower.
BTW, Printrbot had the first dual drive extruder (the Gear Head). Ironically you showed the single drive Printrbot extruder that preceded it early in the video (1:01).
Didn't actually know about the printrbot one. Thanks for sharing.
@@Vector3DP almost no one knows about the Printrbot Gear Head! I have a couple on my Printrbot Metal Plus Dual Extruder that's collecting dust in the garage.
Did it reach europe?
@@Vector3DP no idea! I'm not sure what Printrbot's international marked was like.
Brook Drumm sent me the Gear Heads, so I didn't even purchase mine.
You are measuring how out of round the gears and shafts are. The errors you see are the sum of all of these out of round errors. At high flow you are also getting slipping. At different speeds you get different timing because its literally running slower or faster.
Dual drive is for consistency not force.
Awesome work! I'm here still using one of the original steel extruder wheels from my Replicator Dual that's idk 15 years old now 😅 Printing at 26mm3s to 36mm3s, depending on the filament. Can even print up to ~45mm3s with highspeed filament on the old MK8 style extruder. Volcano heater block with a MicroSwiss CM2 0.4mm nozzle. No thermal paste, i simply tapped the block to ensure smooth matting on the threads and rapped in Kapton tape.
I designed my own extruder arm for the MK8 setup to hold a larger U shaped idle bearing just below the extruder wheel as the filament curls when one side is elongated by the bite of the teeth. The old idle bearings where V shaped and i noticed provided resistance when enclosed and everything was hot. I'm running a StepperOnline 17HE12-1204S 42x30 Nema17 stepper rated at 26Ncm(36.82oz)/1.2A, running it at 1A with a TL-smoother 😅 All this on a heavily modified Creality Sermoon D1, which i purchased open box for $280 some 3 years ago (retail was $1200). I knew it was flawed but it had a massive frame which weighs about 20Kg/44lbs and i saw the potential
Finally, you are my hero. You show how stupid the whole dual drive gear is WHEN it´s the same gear that also has to mesh with the first one and to be adjustable, it´s simply impossible.
You should test the Bondtech LGX to compare if having a very large hobbed gear makes a difference. With the larger gear there will be more "teeth" engaged on both sides than the smaller type. This is very interesting data.
Yes yes! Please LGX Tests :D
A little but it hsouldn´t change too much, you will still get only a line contact even with the larger gear. OMGv2 should be better and cheaper.
Very interesting! I love this all these data videos. I’d be interested in seeing a similar video but with a carbon or wood or similar filled filament that are more prone to jamming/clogging in the nozzle. Seems like extruders that are capable of higher extrusion force might have an advantage their in terms of clearing blocking particles. Also, I’d love to see the addition of a wheel/gear driven encoder to actually measure how smoothly the filament is moving. Do force oscillations always mean more inconsistent extrusion or could it be “smooth” extrusion with oscillating resistance?
Mellow 3D have recently released a version of the Sherpa mini with double helical drive gears, designed to reduce backlash artifacts.
Would be interesting to see it compared with the standard straight cut gears in the same manner as this video.
I picked one of the libra extruders up. It outperforms a standard sherpa on an identical printer.
i wonder how this test would work on TPU. i've never had good luck running TPU on anything short of a dual drive.
Best analysis i've seen so far regarding this topic and force cell experiments. I believe the ripples are caused by the insufficient melting capacity of the hotend; when you get close to the maximum flow at which the material viscosity (U) is "normal" - i.e. the viscosity value at a given hotend temperature at a standstill or at low extrusion speeds, let's call it U0 - you start to have an oscillating behavior where U increases, which makes the flow reduce slightly temporarily, which in term leads to a slightly higher residency time until U0 is reached again, which then leads the flow to return to the previous value. And so on, and so on. (I know this happens from empirical testing) Since you verified that with equal or very similar values for two totally different extruders, that would corroborate this thesis and point to the extruder.
Also, I am of the opinion that dual gears are not needed in any case for regular 3D printing; there are other factor limiting the extrusions. I only use Titan and have never, grinded a filament under heavy testing with 1.75 and 3mm filaments on a supervolcano. I admit that, eventually, having more pushing force would compensate for the under-extrusion that starts happening at very high speeds and layer heights BUT I don't thinks it's the correct way to just try to push harder by brute force in a desire to overpower the physics of the other limiting factors.
Very well commenting. If we look time temperature and time feed force graphs at 18mm³/s, they are probably matching.
Really nice serie, please can you do a variant with the high flow hotends. I will be really interested to see some experiments with retraction
Nice to see more tests!
A suggestion for improvement: 1) Convert the x axis from time to mm of filament. (also works for the frequency)
2) For the Y Axis the amplitude has the same unit (Amplitude of Force in g), right?
3) A clearer separation between Hotend and Extruder would be nice. Extruder alone could be tested by adding a constant resistance at the output, instead of a hotend. Apart from just pressing filament on the load cell, adding a wanted resistance could be quite hard to do I imagine.
I like the idea of testing the extruder without the hotend!
putting a constant force on the extruder is actually pretty easy. just put it upside down and hang some weight on the filament. you could then measure the velocity of the filament and see how consistent it is.
this is one of those brilliant yet simple ideas that make you feel stupid for not getting it yourselves... ;) anybody has done this? do i have to do it? its much more simple and effective than anything i have seen so far! :)
If you pinch a rod enough to deform it, it will lengthen a tiny bit. Ask any blacksmith...
The feed-rollers are indeed pinching the filament enough to deform, which leads to pulsating flow.
Coordinating feed-rate to synchronize with pinch-roller induced pulsation might be a solution.
Fascinating data; the fact that the bondtech BMG performed comparably well with and without secondary driven gear was very surprising and seems to draw into question some of bondtech's past claims re: superiority of dual gear drives. I'm glad to see the titan held it's own well given its age, it's an elegantly simple design that still adorns one of my printers.
I would be very interested in a similar comparison of a number of different aliexpress knockoff bmg gears; are the specs for the measuring rig you're using available anywhere online?
PID frequency? Could just be the pulses of heating or possibly even an oscillation in the PID causing temperature fluctuations and thus viscosity changes effecting the variable force...
Would be interesting to monitor the NTC temp as well as the voltage across the heater cartridge
PID frequency needs to be actually measured on each specific printer configuration, it's not hard fixed, but it should be between 6 and 8 Hz so outside the chart to the right. Correspondingly its harmonics are then multiple of that and also outside the chart. A possible source of visible spikes can then be intermodulation, where for two different frequencies, their sum and difference also imprint.
To be considered whether measurement introduces some sort of aliasing, where for example due to sparse sampling and lack of filtering, higher frequency signals get reflected off the Nyquist and folded back into the visible part of the spectrum on the chart.
I would like to see this testing with the mellow libra mini to see if their design reduces this
Thanks for the video. I wonder what the clone test with the filed gear would look like with the proper bearings present. Also, does having particles stuck to the drive gear going round (like we saw in your tests in blue) affect the results as well? Some of the extruders have wiping wheels to clear these away. Really keen on geared down extruders as every time I try and use non-geared ones, the filament always stops flowing with the print click of death resulting in nozzle jams. Never happens on a geared down drive because of the higher torque.
The clones are notoriously random quality. One of the biggest issue is that hole for the filament is not aligning up with gears and will bind and waste energy to squeeze it through. checking this or reprinting new case would improve quality considerably. I have bought around 10 orders from different suppliers until I found a few ones that provide accurately manufactured extruders with decent gears. I goes without saying that they need to be disassembled and washed because the almost always contain machining metal shards that will destroy bearings very fast. So unfortunately this test on "cloned bmg" is not opening anybody's eyes, because some of them work poorly even just testing the filament moving through extruder and you know it's bad and other ones look and act similarly to original (ofc. these would need this kind of scientific test to see real difference).
Id like to see what some of the "precision" exteuder gears with helical cut gears, milled drive gear, and are millied from a solid piece vs gear/set screw/shaft and uses better bearings on the driven gear (not needle bearings)
Missing a needle roller bearing set is a pretty important factor, and makes your testing data flawed, the gear will apply un even pressure, they need to have 2 sets, regardless if you filed off the teeth.
I would just ask one thing :
What is that pla blob on the titan's driving gear ?
In my oppinion it explains the riples on the extrusion force on higher extrusion rates :
when noting :
Ft the force tangent to the surface createdd by the extrude in our case,
Fp the force perpenticular to the surface created by the extruder's spring
cfns, the coefficient of not-slipping friction
cfs, the coefficient of slipping friction
commonly, cfns >= cfs
on lower extrusion rates, we get Ft < cfns*Fp, specially on the blob were the coeficent of friction drastically decreases, the filament stay synchronised with the gear thanks to the not-slipping condition (Coulomb's friction Law)
on the highest extrusion rates and when the tian's gear presents the blob , Ft tends to exceed cfns*Fp (because cfns decreases drastically : switching form a metal tooth driving it's way in the filament to a soft plastic to plasitc contact parallel to the direction of the filament), thus not respecting the not-slipping condition of Coulomb's law, the forces now respect the sliping equation : Ft=cfs*Fp, wich explain the sudden drop of force on the filament.
It seems that this origin correlates with the adequate frequency (I am not sure not having the original data).
Also Coulomb's solid-solid friction model is simplified, the effects may be better descirbed with Stribeck's friction model
The timing of those peaks make me wonder if that is resultant of heater cycling? What was the temperature variations during the test?
I think if you added a very sensitive scale to the mix it would improve your results. Run the test at a specific distance and measure the weight of the filament that is extruded. This will give you an amount of filament that actually is extruded instead of just the force used. I think it would be interesting to see the amount of force vs actual amount of extruded filament. You could also use this to test filament extrusion speed vs actual filament extruded.
This is beat frequency, when superposition of two higher frequencies close to each other create a new constructive and destructive amplitude at a new frequency equal to the difference.
How does that 18mm^3/s compare to the PID of the temp control for the hot end.
Straight-cut teeth tend to mesh according to the tooth profile: the quality of manufacture of the teeth also impacts this. You'd be more likely to get resonances ("cogging") as the applied driving shaft torque goes up if either of these are less than optimal (think of the noise of a race car gearbox with straight-cut teeth).
I've always wanted to see dual-drive with helical-cut teeth meshing, to see if it makes any difference.
It is noticeable on the Titan images there is some filament stuck on the gears. This might well affect results once per rev.
I have both titan and bmg extruders on my printers. I find that the titan handles flexible filaments far better.
I honestly can't decide which extruder I like better as a daily. I think I maybe leaning toward the titan. You test out the omg or ezr?
So I wonder for an ender 3, that reaches the limit first, if the hotend or the extruder. I have an Ender Neo with Klipper that starts to under-extrude when is near the 9mm3/s mark. I'd like to push it to 14mm3/s aprox and call it a day, but not certain if I need to change both things or one first. I was thinking about a trianglelab BMG 2.0 clone to start with, but after seeing this video not sure if it will help a lot (if the hotend is what is already limiting the extruder for example). Any ideas?
Do you have a bondtech lgx laying around to test?
The extruder is the only thing influencing the force if you try to squeeze more filament through the hotend than it is capable of handling. Not one of the extruders was able to feed 18mm/s through the hotend, but the dual-drive hotends were able to provide a much higher force before the filament started slipping. The big jumps in the force are the points when the filament slipped and are not influenced by the the drive train but mostly by how easy or hard the filament is able to be compressed and the lenght between the extruder and the nozzle. Very hard and rigid filament produces a high frequency and soft filament a low one. A longer path between extruder and nozzle produces a lower frequency than a long path. The test shows that the dual drive extruders are far better than the cheap one, because they "fail" at much higher resistence of the hotend and must have pushed more filament through the hotend than the single-drive one. The spikes in force are not comming from the extruders but from the filament jumping back and at that point this doesn't matter because you will have underextrusion and your hotend is your bottleneck.
I would like to see this test done with more modern extruders like the Orbiter 2.0, the LGX Lite and the TBG Lite, I think the Orbiter at higher extrusion rates and pressures should be smoother due to the planetary gear system, and the others give better grip due to larger diameter hob gears, not many are using BMG's or Titans in direct drive these days, due to weight, the closest we use to a BMG is a Sherpa Mini, which would most likely have the same flaws as a BMG, but with less weight.
I'm looking at those . But Weight reduction seems to be their actual 'real life' advantage .
The big fliament drives , given the above study.. would give NO advantage, only Marketing ones..
To MY surprise... there is only? a 90 gm weight reduction from my BMG /pancake motor /carriage assembly.
Half the physical size certainly .. although not near enough weight loss to warrant the $$ /time /effort AND risk in trying the current Flavor of the Month.
The issue I had with non-dual gear extruders during my pre-direct drive era (currently a ratrig man and pushing fdm to the limits) was that some filaments would simply be 'waltzed' flat and jam in the bowden tube. Bowden obv needs more retraction for a good result than direct does, amongst the traditional debate between the two (there is no "right" answer), and having good tensioning control means a lot as well, but still: interesting videos to watch. I'm honestly intrigued by what we might expect from Prusa's new no-backlash yet also "non-dual gear" extruder. Currently running LGX (the load/unload is just so great), and happy, but always looking for better control.
From the testing I've done so far, I suspect we may not really need anywhere near the amount of idler tension that many of us think we need. Might do a video on this to so we can find out.
@@Vector3DP I remember watching a video by someone showing how the idler tension affected maximum flow, but I don't know how to find it again :(
Normal extruder will need more pressure from the arm in order to get more grip but increase in friction, this sometimes couses motor skip.
The dual drive grips on both side and doesn't require as much spring tention.
As for the reduction gear, it makes it more accurate, and reduce weight since it doesn't need a big stepper motor which is ideal for direct drive
Really enjoying this deep dive into the physics behind squirting out molten plastic. I was wondering if, using your set up, it would be possible to use the force output of higher flow rates to control the heater? The variable of a constant temperature would be more of a hindrance to flow rate, rather than how much the extruder is pushing. I guess what I'm asking is, if your setup is capable of tuning the heater to increase its output based off of the force it sees?
Yes, i think technically there would be nothing preventing that sort of closed loop system. I suppose the real question though is: would it help? Probably a bit.
@@Vector3DP Yeah, I really do think it would help. Given the volume your trying to push through, the heat would only be able to penetrate to a certain depth into the filament. Of course, you could also be reaching the critical point of thermal conductivity of the filament itself, in which case, it wouldn't be much of a benefit to turn up the heat. Guess it's time for a Bondtech CHT nozzle at that point. Cheers mate, keep up the good work!!
Unfortunately, only the pressure on the hot end is taken into account here. However, if you consider the entire system, the resistance of the path from the spool to the extruder, i.e. the pull of the filament from the roll, is missing. This results in completely new aspects when considering the forces required to transport the material. Slippage of the filament can also be caused by an unfavorable material flow between the filament spool and the extruder; this also includes long distances, friction in feed hoses, a tight radius in the material feed, etc.
If the unexplained frequencies don't correlate to the gear teeth or filament 'bite' zones, then the force spikes are either being thermally controlled by the heating elements cycling, or it's the stepper's internal 'cogging' forces being amplified by the gear-trains those extruders provide. I might be wrong but I think when microstepping etc, there are still regions which have more torque to them where the stepper elements can form stronger M fields.
You need to take the nozzle out of the picture and just test the extruder force against a series of constant forces (like a weights hanging from the filament) and record in fine time resolution the filament position so you can derive speed.
I understand the reasons for the methodology but can't see the applicable aim. To what end? So that the amount of force is known? If so the point stands that the amount of force is irrelevant given that the hotends flow capability is the bottleneck on any available forces utility.
@@m3chanist it is a controlled environment to test how an extruder reacts to constant load force. The aim is to test if an extruder achieves constant and smooth force and speed up to the limit load. If an extruder does not pass muster in such idealized conditions, it's already subpart quality.
Look at the gear tooth profile.
The Bond tech is likely using a proper gear tooth profile so the gears are transferring almost a constant force, where the knock off is likely not.
The flow variations I've measured on fibre filled filaments might be an exception where having some extra peak force might be helpful to overcome semi-clogs.
This is true, but if there is any risk of full or partial clogging would you not be much more comfortable with increasing the nozzle size than just hoping the extruder overcomes it quickly when it happens?
@@Vector3DP True, but without good flow or force measurement I don't know how I'd know it was happening. Sometimes you can see it in the final print, but mostly not.
I think the behavior at high flow rates deserves some more detailed investigation. In particular, checking actual filament delivery speed during the test and hotend power consumption. My theory is that the force oscillations are caused by hotend duty cycle, it fails to keep constant hotend/extrusion temperature and switching between high and medium power and resulting temperature swings affect viscosity of the melted plastic. High forces for the higher quality extruders suggest they're not letting the filament slip and show actual force needed to push the filament through the hotend at insufficient temperature while lower forces suggest the extruder slips on the filament and the force measured corresponds more to at which force the slipping starts to happen.
you can call it a myth, but the trianglelab matrix certainly has worked better than my older Titan Aero for my small head printing in ABS. Its hard to flow ABS on a good day, so using a .2 nozzle means its MUCH
nice food for thought, thanks! My conclusions: for regular hotends, more force is not very helpful. For high-flow nozzles such as volcano and the new-fangled CHT ones, higher forces may be a benefit, as there is a lot more viscous resistance there. Also, even for regular hotends, dual-drive is better because we can relax the compressive force on the rollers, so we are less likely to grind the filament if we are doing a lot of retraction in one place.
this might be the motor or driver also due to the high force. also note that 15/16 cube are at about maxing out the most devices melt capabilities (12/13/14 at cheaper clone stuff).
We need videos on how to obtain data and use Fourier analysis on it, it would be great!
When you're measuring the 'gear diameter', to compare with the frequency suggested from the Fourier analysis; is that the outer diameter of the wheels? Assuming that they are involute profile gears, the point of contact between them is not at the outside tip of the teeth, but rather at a 'reference line' that is nearly half way down the teeth. Strictly, the reference line is inside the outer diameter by the 'addendum', which on metric gears is m millimetres, where m is the module number of the gear. That's likely to be 1, for gears of that size, and hence the reference line diameter for those gears will be exactly 2 mm smaller than the outer diameter. To be certain, the module number of the gears would need to be identified, (or the equivalent if they are Imperial, but those are defined in a different fashion), but I think its very likely that this is the case; and hence it really is the gears that are casing the primary waves seen in the force graphs.
Hi! Can you share the stl of the extruder gears showed at 0:23? Thanks!
Interesting. Maybe try a split nozzle with high flow rate on titan? Maybe the resistance could be reduced this way, allowing faster printing while maintaining consistency.
Did you happen to measure the amount of filament extruded? I bet you would see more under extrusion in the clone than the others. I disagree with your conclusion and i think the titan and bondtech are performing better because the likely extrude more. But I also agree with your assessment that even though they are likely extruding more, it is likely not a better result in terms of print quality.
This has some serious implications. If the software measured motor current (which correlates to force), it can slow down print speed when the force is too high. Or dynamically adjust nozzle temp or any number of creative solutions. Also, perhaps thermal contact resistance should be a performance metric of extruders. Filament can have a spec of 95% extrusion accuracy per set flow rate per degree c per hot end thermal contact resistance. This would demystify various hot ends and filament blends!
Exactly. I came here to make your first point as well. Measure the flow rate instead of using the input. For a given hotend and nozzle, you should be able to make a calibration curve and directly determine the flow rate from the force (see math below). With that in mind, I'd assume that the clone extruder just doesn't reach an actual 18mm³/s flow rate. When it tells you it does 18, it actually only reaches what the others do at their 14mm³/s setting.
If you assume that the molten plastic traveling through the nozzle always has the same viscosity (n), you'd expect the force (F) to be proportional to the extrusion rate (v).
F(T=const) = g * n * v
(where g is a term depending on the geometry of the nozzle.) If the hotend runs into the limit of its power output (P_max) (but still heats the filament fiber evenly to its core), then the temperature will decrease as:
T_P = T_room + P_max / c / v
Where c is the specific heat capacity of the filament. Assuming the variation of the viscosity with temperature is sufficiently close to inversely proportional (n = 1/T), you'd expect the force should be proportional to the cube of the flow rate:
F(T=T_P) = (g * c / P_max) * v²
In practice, there would be some flow rate at which thehot end reaches its limit and the force curve will transition from the linear proportionality to the approximately quadratic one. Which seems to be what we're seeing at around 12mm³/s.
It just occurred to me that you should probably specify which PLA you use in these tests, maybe even use Prusament and specify which spool it was since they publish the dimension variability data on each spool.
Why ? This would not change anything resultwise.
Could the ripples be created by uneven pressure from the 2 feed gears (perhaps due to play in the driving gears because of misaligned shafts?) ...
... OR...
could the pressure gears be oblong or different sizes?
... OR... Is the gear holding fixture flexing from the pressure changes due to filament not being melted evenly or quickly enough and perhaps almost plugging the nozzle?
Stepper motor natural resonance making its way into the frequency charts, perhaps?
What is the unit for amplitude? Is it mg?
I think you were exceeding the the hotend's ability to melt filament at the rate you were feeding. I think the load spikes are the filament being too cold to flow well then dropping when it gets hot enough.
Are the massive wobbles in the E3D Titan at the highest flow due to the blob of PLA caught in the teeth that can be seen in the video closeup of the hobbed gear?
Nope, otherwise it would have match up perfectly, plus i'm pretty sure its at the side of the teeth where there is no filament contact
The oscillations are happening inside the hot end due to the phase change and thermal conductivity of the filament. Material density is a slope, but phase change is a binary. A very flexible solid changing to a very viscous liquid may be perceived as linear, but the actual transition from solid to liquid is immeasurable. For a smooth output, the filament has to fully melt within the crucible portion before it reaches the nozzle. Otherwise, you're just packing a solid into the transition between the two. While still solid, it'll just stop at a physical wall.
But, once that solid filament reaches a point where it stops, it's also coming in contact with more heated surface area closer to its center. Thereby "flash" melting a portion of filament, which will flow easily, and make room for the next solid section to close the gap. Repeat, repeat, oscillation.
The 'gap' in melting is due to most plastics being excellent thermal insulators. The heat is only going to propagate over a small distance, which at the correct flow rate should be within the length of the crucible. If the heat outruns the feed, it will lead to a bulging glob building up at the top. Or, if the reverse is true, it will result in very low viscosity extrusion at best and oscillation at worst.
How do you make these animated graphs? They are fantastic
What about temperature fluctuations due to PID compensations causing viscosity changes?
Isn't the point of dual drive to handle more flexible filament and not necessarily flawless extrusion of pla
Great analysis as always
I was ready to drop my clone until you showed the missing bearing.
Probably worth checking if you have two bearings
@@Vector3DP If one bought it assembled, for sure
Could you test the same extruders in a Bowden configuration and compare the results?
If you printed with the same hotend, same filament, at the same temperature, then the reaction force for a given flow rate should only depend on the hotend. The extruder wouldn't matter. As you state in the video. So, this was a hotend test. It also raises the question of how did the clone BMG extruder push the same amount of material with significantly less force? To me, that would indicate better performance, or an error in methodology. Were the same lengths of filament extruded for each test and weighed? The best extruder should have the heaviest extruded sample of filament, meaning it most closely performed as it was "told" to.
3:16 Shifting your 3D Printer’s transfer case from 4wd to 2wd. :)
So basically- nozzle and hotend upgrade is better than extruder upgrade... if both are like CR10 stock..?
Have you taken into account the fact that stepper drivers are also going to produce some measurable ripple? The DRV8825s were notorious for this, and could cause "salmon skinning" on delta printers in particular, because of a huge step jump in microstepping. How have you calibrated your machine to account for these things? You have to take as many factors out of this equation as you can in order to definitively quantize and isolate the drive system itself. Do these ripples happen in a system with a lot of gear-down, such as the Bondtech LGX Lite? What are the effects of using a 0.9 degree step motor at full-stepping to achieve these flow rates, where torque from the stepper should be quite consistant...
Simple answer: Fourier analysis. at the meaningful resolutions that i was working at with this data, and the time steps I have, it really seems every repeating pattern. Obviously there is still A LOT of work to be done with testing many other parts of the system, but If i tested them all before making a video, that would be literally years of work. I'm doing incremental changes and learning about the whole system step by step and updating you all on the process as i go along. hope that makes sense.
Boy, it would have been great to throw the modified clone back on at the end to ensure the result repeated. Just to rule out something partially clogging the hot end.
Also, did you weigh the amount of filament extruded in each test to validate the actual extrusion? Just reporting force during a commanded extrusion rate doesn't tell the whole story. Since you are trying to show a single drive extruder is as effective as dual drive, but without certain artifacts, you must prove the single drive extruder isn't slipping.
I'd love to see a revisit to this test to fill in the gaps! Keep it up.
Some time ago I developed a testing rig for extrusion of PEEK which is quite similar to what you built. But it also features a rotary encoder for measuring the filament slip very accurately, two more load cells and also a PID controller that can be enabled to compensate for the filament slip. My thesis was only about developing that system so I didn't do extensive tests with it. Let me know if you want to read it :)
Hello,
I would like to read your thesis, is there somewhere where I can find it ?
@@geekoid183 It's not published since it's not a PHD so I'll need to send it to you via e-mail. Also you'd have to translate it from German into French. But I think the diagrams are more interesting than the written text anyways :)
Hello! I'd like to read this too. Please send me your thesis, if you can
@@user-ep7ti6de1v Message me your E-Mail address and I can send it to you :)
See if this correlates with extruder temp. Try and implement feed forward on extruder heater drive power. Knowing the heat of formation for the volume of filament being extruded, you can know the amount of heater power required.
Given that there is an seemingly an ideal force required for a specific extruder /hot end pair at a specific temperature...
So, the next question is :
* Without your test setup, what is a deterministic and reliable method to determine that ideal torque and speed for a specific temperature WITHOUT having access to your test rig?
I almost think that question implies that hot ends and extruders should be design as a matched set to get the best performance.
Could you check out the mellow dual helical extruder gear? would be interesting
Could you update this test with the LGX/LGX Lite please?
Note: if it wasn't tested, 'dual feed is better' was a hypothesis not a theory
Dual gear can usually print flexible filaments better, can grip filament better as you saw, and can create more accurate prints. If you want more torque then you would use a geared reduction extruder.
This is wrong. Dual gear grap harder but rhey don´t print better. This is exactly what he has shown here. This also goes for flexibles, since a harder grip means simply more deformation of the material.
I think Bruce needs a mustache.
Also, you should really test a DyzeDesign extruder, they use 4 gears....and Vez3D has a very interesting hot end doing +130mm^3/s, that could be a good match up with your testing system
I wonder if the force oscillations could be mitigated with a modified hotend control strategy at higher flow rates