What I've learned about dual gear extruders and how to patch them, maybe

[@MihaiDesigns]

Check out my newly started Patreon page if you want to support my work www.patreon.com/mihaidesigns

[@l3d-3dmaker58]

I have a similar issue but the pattern is completely random, as if there were extrusion inconsistencies, I'm gonna test either .9 motor or gear reduction, it may be rounding errors, I need to print your test pieces!

[@lazyman1011]

Nice work!

[@KLP99]

I remember seeing another video where the maker figured out that changing some Marlin setting to match the number of gears the extruder had removed the wood-like pattern. But as I recall, it was done for a single gear, not dual geared extruders. That it can be done at all would indicate that someone smarter than I am could figure it out... Go for it!

[@lio1234234]

I guess I must have just been lucky with the tolerences inside my prusa extruder and drive gears then, I've never had any surface patterns on my prints.

[@hillfortherstudios2757]

I still dont understand what you are auggesting to solve this issue on a mk3s. Please clarify.

[@poonampadhi658]

Well if we use wood filament then we get free wood texture 😅

[@MihaiDesigns]

😂 indeed

[@randgrithr7387]

Based

[@BlackDreaded]

just had a look for possibilities like this :D looks like I have to get dual gears for my ender3 😂

[@absalomdraconis]

@@BlackDreaded : Another possibility of course is a bit of patterned "noise" in the extrusion distance.

[@goury]

(= But there are better ways to get it

[@marsgizmo]

Your research and findings are super well detailed, amazing job! 👏😌 Let’s see if someone is stepping up to improve the gears ⚙️🙂

[@jamespray]

Seems like there's an aftermarket opportunity here, doesn't it? Drop-in replacements with herringbone gears ... I'm not brave enough to try a hardware startup even if I had the expertise, but there's gotta be somebody ...

[@user-yk1cw8im4h]

why do you have the exact same logo as mikai

[@absalomdraconis]

A thought that comes to mind is driving both gears off a common shaft, via a bevel arrangement. There might be another type of noise produced by that though.

[@Bordpie]

@@absalomdraconis Cylkro gears could be used for that. They are a special type of angled gearing where one of the gears can be a simple straight involute gear. This means you can have a 90 degree gear mesh where one of the axles can translate from side to side a bit (like in an extruder) and have zero backlash or affect on the gear meshing. An example use case I saw was for a rolling mill, where the rollers can change position to accomodate different thicknesses of material, but there is no affect on roller rotation as the rollers move up and down (essentially just like an extruder). I believe they can be cut with the same equipment as standard involute gears, but they are quite rare and expensive (only made by one company really) since they have niche uses.

[@Bordpie]

Standard involute gears transmit continuous smooth rotation even when the centre distance between the gears changes e.g. if the gears are close together when driving a flexible filament or further apart with rigid PLA, the gears with both rotate at constant speed, at least in theory. In reality you will have a bit of angular misalignment between the axles. This produces meshing error, increases noise (although not really relevant in this situation) and creates edge loading where only one end of the gear meshes while the rest doesn't carry any load. I am guessing this is what is causing these problems, and would explain why some extruders of the same design are having this problem and some are not. The ones which do not probably happen to have very good shaft alignment. Helical and herringbone gears are no different in this respect and herringbone gears are much more expensive (and actually less efficient than spur gears but again this is not important in this situation) to manufacture accurate metal gears using specialised forming machines. A potential fix could be crowning the spur gear teeth. This is a common procedure (although adds to the cost) for compensating for misalignment and reducing noise and meshing error. It basically makes the tooth face slightly convex instead of straight, so when the axis are at a slight angle, the contact is still in the middle of the tooth. This does reduce strength slightly but again is negligible in this case.

[@fabianfenner3067]

I had this pattern with my Ender 3 V2 which has a Microswiss Directdrive upgrade. everyone called it inconsitent extrusion or blamed my filament (ASA Polymaker), frame or settings. I was troubleshooting for a long time with no success. Yesterday I gave this a shot and the results are crazy. Thank you for this research, I‘m really grateful for that. I hope this comment may help a view viewers with the same issue. Unfortunatly this video didn’t pop up 10 months ago because of the stupid youtube algorhytm… Thanks and have a nice day

[@BoOmChAkKaLaKkAxD]

05:18 Saying "it wasn't that hard" After changing literally every perimeter is very brave. Had to LOL very hard. Thank you for the interesting content. Greetings from Germany.

[@MihaiDesigns]

Haha indeed, I didn't think of it like that, but you're right 😂

[@goury]

It wasn't that hard compared to the whole RepRap development to this day. It was much harder though than just mashing gears without understanding very basic mechanics and geometry, like these wannabe-engineers did and now are copying the same bad design.

[@TheGoodandBadJuju]

Hard is only four letters blocking restricting your view of how simple it really it is. 🤞🏽❤️

[@Agnemons]

You could solve the meshing issue by adding two idler gears, you need two to maintain correct rotation direction. If you add one idler directly above and driven by the motor driven gear then the second idler at 90 degrees to the first idler then the second feed gear at 90 degrees directly below the second idler (power goes up, left (or right) then down). The carrier frame for the second feed gear would pivot on the axle of the second idler gear. Ideally the idlers would be slightly larger than the feed gears to ensure there is no gear clash with the feed gears.

[@RockIsLife001]

Bond tech needs to give this man a job. So does prusa.

[@tylersutton2216]

ikr

[@goury]

Avoid working for Prusa, seriously. He pays jack shit and his work conditions are literally toxic (employees are forced to inhale all kinds of fumes from their factory for the whole work day).

[@tylersutton2216]

@@goury wow

[@goury]

@James McCaul almost. I've rejected multiple offers from them. You know this feeling when you're a senior software developer with 10+ years of experience in development, leading, management and architecture and they offer you $3k gross salary and the contract says they'd own you and also can dump you any moment they want and you want to just get to their office to tell em "fuck you" in person? Prusa was not the one to teach me about this feeling, but he sure did reminded me well on it. As for the work conditions, you can watch their office tours made by their advertisers here on youtube. Those who pretend to be hobbyists, but actually advertising him rather than doing any hobbies.

[@SkilllessFlorid3r]

I don't know, this is the first time I heard something like that.

[@Longestnickevaaaaaar]

I really appreciate the amount of work that went into this video. And I also love how it is presented. Thorough, yet intriguing. Well done!

[@JAKON72]

This video is incredibly informative. You should add 'salmon skin' and 'moire' to this video's tags so it can be found by people who are experiencing this problem with their printers.

[@MihaiDesigns]

Good point!

[@jeremybunn72]

Initial experiment with the X idler door... It is working! Huge difference on a key part I print allot... WOW!!! Printing more, allot more of these now. Thank you for staying awesome!

[@MihaiDesigns]

Thanks Jeremy, if you have time to fill in the survey for the X idler door it would mean a lot.

[@everydaygear3730]

It's not cutting costs, it's keeping costs down - which does NOT always apply to the R&D phase. Software development is no different, especially when it comes to "scale" - in your world it's practically free (server space and maintenance, and software updates) versus tangible products with traditional costs (material, fabrication, labor, shipping, etc). Great video! That comment just stung me hahahaha

[@ManiekFPV]

Mihai thank you! The spring helped with my inconsistent extrusion issues. Walls are now even and smooth. 😊 I didn't have it before, I don't know why I didn't put it on.

[@shaunmorrissey7313]

Fascinating, looks like I was right to stick with titans 😀. Seriously though, outstanding detective work.

[@ErosNicolau]

Thanks for taking the time to do the detective work on this one! I personally am never bothered by these details, but know a lot of people who are and will be happy to learn there's now a known cause for this!

[@MihaiDesigns]

Glad to help! Indeed, only some people get bothered by these luckily. I became a bit obsessed about it 😅

[@demianmoody148]

10:55 that's probably runout on the extruder wheel, meaning that the part of the extruder wheel that grips the filament is not concentric with the axis of its rotation, which effectively acts as if the extruder wheel diameter increases and decreases repeatedly to create that streched out wood pattern

[@MihaiDesigns]

Oh I see, that could be it!

[@MadByte]

That's what I was thinking. If there is any slop in the gear's inner diameter the set screw will push the gear out of center. I would imagine this pattern would be doubled because of the two gears. Would it be less pronounced if the set screws were oriented the same (or opposite) relative to each other?

[@soundmindtv2911]

Some mechanical engineering insight - spur gears are designed to be constant torque + constant velocity, but this consistent transfer is only true at a set depth of mesh. The teeth must be machined precisely, with the correct arcs, and the center-to-center between the exes needs to be precise and a fixed distance to maintain the exact mesh. Any variance will cause the driven gear to move at a different velocity than the drive gear, depending on the angle of attack - in other words, as each tooth engages, pushes, and disengages, the velocity of the driven gear will decrease, increase, then decrease again.

[@Doktoreq]

@@soundmindtv2911 If i'm thinking correctly to fix or minimize issue we have to have more teeth on the spur gears - the meshing will be more consistent with less speed variation (the speed variation frequency goes up and amplitude goes down). To have more teeth we either make teeth smaller - which in this case is impractical because it limits the filament diameter (too big and gears are out of mesh entirely), or make gears bigger. If we make driving gears bigger then the hobbed gears need to end up bigger too. Which means that we end up with something looking like BondTech LGX. And now I'm interested how that's performing.

[@jamespray]

@@Doktoreq I ran the test prints with my LGX and did not have the wood grain. I'm an ME by education (though not by trade) and my theories as to why the LGX might not display the same extrusion periodicity as designs with smaller gears align with what you and SoundMindTV mention above (especially mesh issues caused by variable center-to-center), but I suspect the additional surface area contacted by the larger-radius hobs is also a factor.

[@RichardBronosky]

I thought this was going to be an issue the first time I saw this design. I'm both glad and saddened to see your analysis here.

[@billyjoe3309]

Im telling you, when you release your swapable extruder, I will get it, even if you sell them or the plans for them. I'm getting it

[@MihaiDesigns]

Deal!

[@soundmindtv2911]

+1

[@NicksStuff]

@@MihaiDesigns You could create the Mihai3 (Mi-i3)

[@goury]

@@MihaiDesigns make sure to stick to the RepRal License. Don't violate it like Prusa and many others, please. We've been working very hard on this tech not for some businessmen to parasitise on it.

[@brandonallen2372]

@@goury How has Prusa violated the RepRap license?

[@Ernescme]

Might be a bit complex, but you can put the secondary gear on a pivoting arm and drive it with a belt that has a tensioner. In that way you can change the gap and drive both gears in sync.

[@court2379]

Looks like helical gears would even out the variation and eliminate the problem. You don't need to have the herringbone gears though as they will cost much more, and having a thrust load is OK. Either that or just stack two helical gears with one flipped 180.

[@davidson8939]

the issue is the size. what needs to happen is the gear mesh and the filament pushing has to be independent of each other. a more tolerant gear can help - helical, but i suspect what will happen is the problem will become less consistent. I spend too much time doing cad work for my job for it to be fun. I think it can be solved with 4 gears but I will wager that the resulting mechanism will be too big to work. 5 gears may help by getting the stepper out of the axles. these gears are already fairly small for metal gears.

[@court2379]

@@davidson8939 Yeah ideally you put the gears at their proper distance apart, and use some movable coupling to the filament drivers that minimizes any backlash created. Using a u-joint, cv-joint, oldham coupling, or many other designs. The helical gear would probably deal with most of it though without the extra cost, despite not being perfect. I guess you could also just use two smaller motors and drive each wheel separately, synchronized by the electronics.

[@BANGMAN009]

Libra mini

[@janwaclawski6658]

Excellent investigative work. Food for thought and I really need to see if I can emulate these findings this week. It does fully fit the "i changed the extruder hob/extruder type and it went away" reports I've seen over the years, but lacking a real explenation on the "why" and "how" of the matter. Thank you so much for sharing.

[@brucebelvin2058]

You talked me out of getting a dual gear. A single gear has no backlash and works really well with the proper tension.

[@ricktm79]

I resolved this on my direct drive dual-gear extruder for an ender 3 by adding four dampener silicone o-rings (commonly used on mechanical keyboards) between the extruder stepper & its mount. This gives the stepper motor's shaft a little bit of wiggleroom if it needs a slight angle for various filaments. screws can be tightened to adjust for stiffness/more angle, while letting the idle gear stay at a fixed angle.

[@taracair]

Would you mind sharing the details of the idea? Any photos?

[@Zzafari7]

this is not a 3:1 dual gear extruder tho right?

[@lanoxx8075]

Your solution with the double gear was my first thought. It is a common problem in mechanical engineering.

[@goury]

Thankfully, it's not. Gear matching is present in any decent piece of machinery. It's just all these extruders are not it.

[@HungrysitesRu]

Wow, just wow, so much effort was put into it but noone found this before!

[@beaconofwierd1883]

I’ve always been sceptical of ”semi meshing” the extrusion gears like that, good to know I wasn’t worrying for nothing :) I wonder if connecting the gears through arms such that they mesh perfectly would solve it :)

[@genioee]

I got kinda frustrated with the number of teaser videos, but this is one very nice finding and good investigation. I would argue, to keep people interested you can also find better ways to engage with your viewers. The gears are not ideal, also from the perspective of size, as the LGX has spotted. It would be super cool to have another way of addressing both these improvements in a quality extruder gearing system. Nice job!

[@ajosepi1976]

Just a thought: Set the driven gear in a pivoting housing where the pivot point is biased to the gear teeth. Spring tension will hold the gear mesh constant and allow adjustment in the filament drive end of the gears. This will cause the gears to wear out sooner, but should help with print quality. Good video.

[@agepbiz]

Interesting findings!

[@Vez3D]

Nice digging!! Thanks for sharing this. Lots of work here man. Good work

[@MihaiDesigns]

Much appreciated

[@ottovp]

give this man a job and a raise already

[@monev44]

I noticed a pattern in my prints that changes alignment pattern with perimeter geometry, I thought it might be something in the roundness of the gears in my bondtech geared extruder, but I was hesitant to break it open. This makes it seem very likely, and since I know I can't modify it Ill have to learn to live with it at least knowing what's going on. Thanks for disassembling your extruders so I don't have to.

[@shannim5]

Awesome research! I will need to print your STL files and see if I'm getting the same thing. I recently converted my Creality Ender 3 Pro to direct drive with the stock single gear extruder and then started to see surface patterns that weren't there before.

[@MihaiDesigns]

Go for it!

[@shannim5]

@@MihaiDesigns I'm wondering if eccentricity of the drive gear is at play. The gear has two grub screws, but I swapped the extruder and z motors to maintain full X travel. The z motor shaft doesn't have a flat spot like the extruder motor does.

[@3DPrintSOS]

Love watching these. Your work is amazing and the effort you put in to polish everything is admirable. Well done!

[@MihaiDesigns]

Thank you! Cheers!

[@tomtaylor135]

Dual Helical gears with filament teeth in the middle of the gears?

[@gamerpaddy]

maybe try building the fuselab extruder drive system, it doesnt use gears, its basically a tapered nut (they use 3 angled threads or something) that rotates around the filament pushing it like a screw.

[@MihaiDesigns]

Does it still exist? I haven't heard anything in over a year.

[@martylawson1638]

I've seen one hobby build replicate it. The rollers need to be inclined to form a hyperbolic pinch zone. Groved rollers are probably optimal, but fine pitch threads should work nearly as well. Would need some work to find the right roller angle to make a 3 or 6 start thread on the filament.

[@gamerpaddy]

i tried making a rough concept using modified m8 threaded rods slightly tapered. that are static and pretty rough. i think the biggest problem is filament twisting itself. not tested tho. www.thingiverse.com/thing:4907745

[@Bordpie]

I think that is patented? I could be wrong about that.

[@gamerpaddy]

as long as he doesnt sell it? who cares.

[@williampowell2078]

Good analysis. Gear manufacturer here. Simple helical gears would be fine since there's a negligible axial load on the driven gear, so a herringbone gear is an unnecessary expenditure. I will look into helicals at that pitch diameter. Your analysis is spot on. If a spur gear pair is not meshed at the pitch line, there will be a variance in the angular rotation rate of the driven gear.. this creates several problems in the extruder. Picture the cyclic forces on the filament itself. There is a smearing of the filament as it is forced through the extruder teeth due to the teeth failing to rotate at the same pitch rate. The problem here is that the top hobbed 60 degree worm gear that grabs the filament has a pitch diameter that is too large. I'm going to guess that the 42% of the machines issue is a manufacturing tolerance feature for different batches of gears.

[@Gmaniacable]

straight teeth will allways "step from tooth to tooth". hence the angeled teeth in a vehicle gearbox. citroën used to glue 2 together to eliminate axial forces. and offcourse, electrical engines never realy spin perfectly "round" they are limited to the windings inside. adding another gearbox can reduce this. but all this is very deep into the basics of any printer. amazing how you found, tested and work on solutions.

[@NeoIsrafil]

Thank you for finding the problem, the solution is to drive the two gears separately and at the same rate, instead of relying on them interacting. I've got a design in mind but .. no time to prototype and build it :(

[@JoeMalovich]

Two potential solutions I can think of: a gt2 belt drive connecting both sides of the extruder but the routing is complicated by the necessary reversal of direction. The other is to add a second steeper motor for the other drive wheel. This adds weight but it's not too bad if both motors are half size.

[@JoeMalovich]

I was thinking about it, after some googling there exists double sided gt2 timing belts. This negates the reversal difficulty while only requiring the addition of a return pulley which can provide belt tensioning independent of extruder tension.

[@robinte98]

What about making the gears larger so that the distance to the filament is higher and rotating the second gear around the one on the stepper until it squishes the filament? The gears would always be close together and every filament can be used. I don't know if I explained my idea clearly, but in my head it's a simple solution.

[@Dmitriy125125]

Is it possible to solve it programmatically? E.g. if each printer has static deviation from ideal extruder then it might be possible to make extruder motor rotation adjusted to that deviation to compensate it. Some kind of coefficient, like the LA one. The tricky thing is that printer might has to know somehow the position of the gear on a moment of starting printing.

[@MihaiDesigns]

Not very easy. You could in theory do it, compensate for a repeating pattern. The challenge is figuring out the pattern offset, so how to align the compensation with the pattern. I guess in theory you could manually align it using some test print and then never turn off or reset the printer so it knows the pattern offset at all times. Include that in the firmware and it might work. This also goes for other patterns as well. With vibrations it's easier because you can use an accelerometer to make measurements. Well, technically you could add a very precise encoder that reads the filament movement (even though that would have its own moving part that can add error) and use the readings to figure out the offset (and potentially figure out the compensation function automatically).

[@Dmitriy125125]

@@MihaiDesigns if gears issue affects only the speed of filament movements (by making it unstable) then we can make a calibration movement before print starts (e.g. on a nozzle prime action) and measure speed with it deviation. Then having information about speed deviation we can identify pattern offset. To do it we might make a device that can read a speed of filament usage. What do you think?

[@gcod3d161]

@@Dmitriy125125 I’m designing a printer that uses multiple *independent* print heads collaboratively printing an object, so reliability is key and this is something I’m of course interested in. I have a few ideas for low cost reliability, one that I’ve thought about to increase reliability is having a laser on the underside of the nozzle to measure actual extrusion and measuring actual object position/ layer width. That’s the device you’d want to measure filament extrusion, whenever more filament is pushed through the nozzle you could increase the speed of the x/y axis to smooth it out and slow down when the extrusion decreases. I don’t know how that can be implemented in firmware so maybe a god sent software engineer can chime in on that one

[@benm4784]

Great video! First resource I've seen detailing this issue. I've been seeing this exact problem on my old solidoodle, and always credited poor E-step resolution and have intended to eventually replace the whole direct drive extruder. Fascinating to see other factors that can cause this, maybe worth more investigation!

[@Javii96]

Wow that was such an intelligent way to diagnose the issue.

[@themakegoodcollective]

Brilliant work figuring out the problem. I'm no expert on gears but there are several things that might address this. One of the spur gears can be split and rotated. This rotation can be spring loaded. Helical gears may fix the problem and one gear can be moved in or out along the shaft to take up the play caused by different filaments. One of the gears could be conical and axially loaded. Again great work. No the gearheads can jump in and tell us which type of gear we should use.

[@NicksStuff]

You know your extruder's tensioning screw...what happens if you use two? One at each end of the axis of the idler gear. Shouldn't that allow you to minimize backlash *AND* keep pressure on the filament?

[@NaterNorris]

It looks like more of a design flaw. If you are using retraction then when you retract the play in the gear mesh will cause headaches...

[@1tibbers]

You will end up with a worn out pin or bearings in the gear that uses a pin.

[@NicksStuff]

@@1tibbers That's how it works today, I'm just proposing to use a spring at each end rather than just the one

[@Kalvinjj]

I think I get what you mean, two tensioning screws could then allow some twist/rotation of the gear itself, and yes this could solve/alleviate the issue. But I bet you would need quite a lot of calibration work to do manually for each filament.

[@NicksStuff]

@@Kalvinjj Yes, that's the idea. The tension would be set by the springs, not sure calibration would be complex

[@mariokuntzag1237]

Du bist einer der besten Köpfe die ich seit langer Zeit gesehen hab! Bitte mach weiter so!

[@ErtsenPlayGames]

you like 30% right in this video ....the real problem are that gears arent press fit = any force aplied can make them twist etc if the gears arent tight on the shaft = impact the print quality taking out 1 gear stop them from moving all the time on shaft yet they still wont be perfect the new bondtech gears have press fit shaft , gear for motor etc and the quality is ...daaaaamn good

[@stevenlittle5372]

Good findings, though for me the value of duel gears greatly outweighs such a drawback.

[@Thomllama]

Drop Effects OmniaDrop extruder addressed this years ago.. :P since I got my 1st one and discovered it's improvements, all my printers now use the OmniaDrop!

[@MihaiDesigns]

How was it addressed? I'm not familiar with this extruder, but does seem to be using the same gears.

[@Thomllama]

@@MihaiDesigns they are actually pitched at an angle, think like 12 degrees but dont quote me on that. Keeps the gears from separating when inserting filament as badly as normal inline gears so the backlash is lowered a good bit.

[@Thomllama]

@@MihaiDesigns here is my interview with Max from DropEffects and his explanation of why/how he pitched the gears- kzbin.info/www/bejne/jYHaiYl9gKuEa7s

[@VinayVarsani]

The dual helical design definitely looks very promising if the filament can fit through the middle without the teeth slipping!

[@ExplodingWaffle101]

i didn’t think of it like that, but now you’ve pointed it out it makes me wish i owned a lathe haha

[@alessandrogiuliani2915]

Wow…. This is a person that really want good stuff, and he want it to be reliable. Thanks for the awesome work that you are doing

[@MihaiDesigns]

Thank you too!

[@lukepjsmith]

love the work you've done here, just throwing it out there that I've had this issue on a bulldog extruder in the past and it appeared to be a firmware issue.

[@capnthepeafarmer]

From a mechanical design perspective to make the dual gear more effective you would want to actually add another gear and change the geometry. You want to have the gear meshing be invariant of the filament diameter. It's difficult to explain in words, but I will try. Think of a scissor with a central pivot point. The drive would be in the scissor pivot point, then at the tips of the scissor is where the rollers for the filament would be located. The drive gear could be connected to the filament rollers by a timing belt or more gears, but the main point is that as you open and close the "scissor" the gear meshing stays constant through the range of movement. Ergo, gear meshing becomes independent of filament diameter and variation. From a high level that would potentially solve the issue.

[@MihaiDesigns]

Would take 4 gears to preserve the rotation direction.

[@Storebrand_]

I had basically the same idea! you have two belts, 4 belt timing pullies, dual drive gears and two gears and four working shafts. Two master shafts, two slave shafts. The two master shafts is the motor shaft and a static shaft, with the two slave shafts one is static and mashes with the motor's shaft, the other is a pivoting slave shaft which is connected to the first slave shaft and pivots around it. The motor master shaft turns the static master shaft with a belt the motor master shaft also has a gear that meshes with the static slave shaft's gear, the static slave shaft is then connected to the pivoting slave timer pulley with a belt. Essentially it would look like this when closed | | then with the pivoting shaft open it would look like this | /. I think this solves the issue but would be quite a bulky extruder, unlikely to be practical for direct drive. Actually in finishing writing this out, I've already thought of another solution that requires only one belt.

[@azinfidel6461]

Google Zero backlash spur gear, consists of a spring tensioned double gear to replace one standard gear.

[@0LoneTech]

Which gives you low backlash if your force and acceleration are also very low, but doesn't really affect periodic error. It would only affect retractions.

[@Machiflores]

First of, great video. I have also watched all the inconsistent extrusion solutions and none worked. And also looked at ender 3 print that looked better than my prusa mk3 prints. I left the search for a solution as prusa research called it non existent and I supposed people just accepted it. Second, a good solution to avoid inconsistencies in gear movement is to have different teeth count in the gears, usually prime numbers. It might help here. Now you NEED the same tangential speed on both gears! So you either change diameters of the hobbed parts with the teeth or you add a train of gears (an idea I read below) with one being on the axis of rotation of the door. Third, prusa nextruder brought other videos that explain possible issues in low diameter hobbed parts. For example, excentricities. This might be the case of both single a dual gear IE. Lastly, my vote is for belted gears. Might not be the best for traction, but it would be better than no dual. I think it could handle well different pressures of the hobbed against the filament. I dont remember any implementation of this.

[@someoneonly]

This is a very insightful examination into an issue i thought about but never troubleshooted

[@bobo-cc1xw]

As a mechanical engineer showing a herringbone gear and say "it is just gears man" trigger me so hard. Good investigation

[@bryanst.martin7134]

This explains the purpose of going to the new style triple rotating helical rollers to drive the filament axially. Seemed like a gimmick until you explained this. Thank you very much.

[@sigterm15]

I absolutely LOVE your work and your style of thinking - exactly same as mine. All my prints look like trash after moving from PrusaMK2 style extruder to MK3 & dual-drives. Only thing that "helps" is to damp the extruder by bowden or to overload it with high speed, so it is not able to pronounce those oscillations.

[@ottoolsen9676]

it is possible to put them inline with gears between or to extruder motors and use ball bearings to clamp the fillament this solution is used on almost all welders. there are several problems with this gear setup, wear under the gear over time, and wear on the pin and bearings, and the bearing get dry after a while so regular re lubrication is needed.

[@soundmindtv2911]

Troubleshooting at 100. Thank you, my friend. Now we have a starting point to engineer something revolutionary 👍🏼

[@soundmindtv2911]

Ultimate solution: a pellet feeder with a screw extruder. Eliminate filament altogether, and move to pellets. Then mix your own formula, including colors, without needing to stock full rolls of specific composites and colors. Natural base pellets and pigment pellets 🌟

[@gangleweed]

I'm new to 3D printing but many years ago I designed a wire feeder for a machine that fed 1.6mm diam wire into a machine for making plugs for Tig welding gas turbine blades. The turbine blades were for Rolls Royce jet engines and after the blades were investment cast they had small holes in the ends to vent them.....these holes had the plugs inserted in the ends and then Tig welded closed.......production rate for the machine making the plugs was 80 strokes per minute ......80 X 60 = 4,000 per hour. The point is, the rollers feeding the wire were plain grooved without the gripper grooves like the 3D printer filament feeder in question and only one roller was driven the other roller was spring loaded against the wire. Perhaps the cyclic gripping force of the gears and/or the roller griping teeth is causing the pattern that is on the prints.....I don't know, it's just an observation. One thing is also apparent....the smaller the diam of the roller the greater it will indent on the filament surface.........so if the roller(s) were a larger diam the surface area would have a greater gripping force and a pair of rollers in tandem would also have more grip to feed the filament without having to squeeze it so tight. I would even go so far as to say that if a set of 3 pairs of smooth rollers (grooved but no teeth) in line were used it would smooth the flow without having to grip the filament so tightly......anyone who has press tool experience will know how a strip feeder works.

[@MihaiDesigns]

Indeed I was also thinking about multiple in line sets of wheels.

[@gangleweed]

@@MihaiDesigns 6 Grooved wheels in pairs......you would have to couple the wheels together in pairs to get the benefit of a multiple drive, that means the wheels would have to rotate in the same direction for all pairs........perhaps they could be coupled with an O ring as a drive belt to each one on one side to ensure they all rotated the same way while gripping the filament.....the opposite wheels would just be spring loaded to apply the grip........I think the diam of the wheels should be larger than the current drive ones........about 20mm diam and both grooved sounds right for a better grip without actually indenting the filament.

[@BoOmChAkKaLaKkAxD]

What amazes me is the fact that I got 2 printers with direct drive bondtech extruders. One does have the exact problem you are discribing. The other one is printing perfectly fine without any surface problems. While the results you found make sense to me, I just don't understand why one of the extruders of mine are working fine while the other one does produce strange surface patterns.

[@MihaiDesigns]

I ask myself the same thing. I know many Prusa printers show this, but many work just fine. And throughout the data I received from users the same printers and extruder models can show different results. My only guess is that there's very little room for error and any misalignment would result in these problems. This would explain the printed extruders which won't be exactly the same due to variations in the printed parts and perhaps how tight every screw is. It's just a theory.

[@Bordpie]

Yeah, my current hypothesis is it is caused by angular misalignment between the shafts which cause meshing errors. Hence why it occurs on some machines but not others. The fix in industry is tooth crowning so that may be worth investigating.

[@tiagopviana]

Nice discovery. The first extruder that comes to my mind, that probably doesn't have this problem, could be the Bondtech LGX Extruder. Although, it still uses 2 Gear to mesh the 2 filament grip wheel. But because the gears are bigger, and have more teeth, you can get again the pattern but with a small dimension...

[@MihaiDesigns]

Yes, I want to try the LGX.

[@123bookra]

you can buy just the gears or buy orbitar extruder ₪ 62.66 13%OFF | trianglelab inner diameter 8 mm Drivegear kit dual drive gear extruder kit Cloned Btech upgrade for Prusa i3 Bowden Extruder a.aliexpress.com/_vlqHZn

[@minosg]

That's an interesting deep dive . I was wondering though how much it is a bigger issue in direct drive extruder compared to bowden. Fractions of 1.5mm extrusion artefacts could easily be hidden in the bowden noise.

[@MihaiDesigns]

They appear to be smoothed out by the tube. Doesn't mean it doesn't cause other issues, but it does smooth this one out.

[@airsoftfarmer]

I wonder if using a toothed belt with a tensioner to drive the other gear would help?

[@ExplodingWaffle101]

just saw your last video in the recommended- glad you’ve just uploaded 😂

[@3DPrintBeginner]

Excellent work. Can't wait for more content from you!

[@StrikeEagleCC]

Great work! The first video of yours I saw was the "3d printing does not have a speed problem", and I wrote you off after seeing it. But I'm so glad KZbin suggested your last two videos! You have obviously put in a lot of time and effort into investigating and documenting this issue, and I am grateful for your efforts and that you shared your results and conclusions. Thanks! Also, I wonder if replacing the spring tension mechanism with a fixed positioner like the LGX does would have an impact on this. The thought is that the force imparted to the idler gear may have a component that compresses the idler spring, and that component varies with the relative position of the driving and driven gears. A rigid idler mount would prevent spring compression that results.

[@MihaiDesigns]

Haha yes, my most unpopular video that will forever hunt me. I'll leave it there as a reminder. You could be right about the LGX design. I have yet to try it out. I guess to simulate a more rigid tensioning system we can skip the grub screw spring. But I somehow doubt that LGX isn't using some sort of springy solution.

[@StrikeEagleCC]

@@MihaiDesigns I'm not sure about LGX, but I know if at least one instance where the spring has been omitted from a dual gear setup with good results. However, it was done for part clearance reasons, and not in the context of the issue you're investigating here. That is to say, it's possible, and may be worth testing in this context.

[@ferrumignis]

How would a fixed positioner deal with variations in filament diameter? It could slip or bind up.

[@StrikeEagleCC]

@@ferrumignis in hindsight, I wish I had said "fixed position idler". You're right that it could potentially slip or bind if the filament diameter variation is extreme. In my experience, a typical extruder bites into the filament much further than the typical filament diameter varies, so I don't think that would be a problem. I have actually been running an extruder with a rigid idler for more than a year now with no issues. As for why it might help with this issue, it seems plausible to me that a spring loaded idler can move away from the drive element under some circumstances, allowing the filament to "ride up" on the knurls of the drive element and changing its effective diameter. Would be interesting to see tests to verify whether this happens, and if it does, whether its effect is negligible.

[@deathcube2006]

Theres a sherpa mini extruder from mellow that features helicoidal gears, so they did listen to you! Good job

[@MihaiDesigns]

I know the Sherpa Mini but the one I see on the Mellow store doesn't seem to have helicoidal gears.

[@juanmf]

Decoupling the filament pressure from the gears biting each other is the way to go. The filaments biting part should be flexible, so it deforms inwards as the filaments lasses but the gears stay in place.

[@flyingby3703]

An idea would be to use a very small belt with a tensioner. But you would have to redesign the gears for pulleys.

[@InspGadgt]

2 things come to mind...this could also be caused by the friction as the teeth meet and not so much the backlash. At either rate a very thick and sticky grease on the spur gear portions could help with both the friction and the backlash. As for an alternative drive method you could put the spur gears on another axle and belt drive the hob gears. In this method the gear mesh of the spur gears would be constant and you could set how tight they are much like an RC car does between the pinion and spur gears. The hob gears would then be on their own axles and move together in opposing arcs. So long as you make the pivot point the spur gear axle then the belt tension would remain constant. It's quite a bit more moving parts but would minimize the effect you speak of.

[@SirKittyMcMittens]

Thank you for this video! I've been chasing this issue for so long now. Using the extrusion test stl you provided, I get that woodgrain effect. I turn my dual gear setup to a gear and pully like you did, and I still get the issue. I look at my extruder motor, and I can visually see it doesn't spin smoothly, thanks to a knob I installed, which has the unintended effect of being a great visualizer of the motor movement. I personally believe it's going down to either my TMC2209 drivers, or my stepper motor's QC, seeing that I messed with everything I could. We'll see if I find the root cause of my problem.

[@TheCebulon]

So, use a belt to drive the second gear. And a belt tensioner. That does not change the length of the belt when the second gear moves due to changes in thickness of the filament. Will be complicated, but the tension on the filament will be the same during the print.

[@brunoosti1]

Single gear with toothy idler? Squeeze that spring as much as your filament can take and have a middle ground in the triangle? The Sidewinder x1 (no pattern) have a grooved idler, but that sill could slip. Excellent investigation work!

[@MihaiDesigns]

Yeah, I usually tend to squeeze the springs. Did the same on my Sidewinder. I can barely load new filament in 😂

[@brunoosti1]

@@MihaiDesigns on my old Anycubic i3 i used to tighten the spring too much, leading to under extrusion (because of the gear biting into the filament). After some testing, 2 complete turns seamed to give the best grip without hurting the filament. At the speeds we print with these hobbyist machines I don't see the need for dual gear extruders, still.

[@arturofernandez8487]

I’ve just installed a hemera on my cr10, I just hope it’s not as bad as this, but at least I know what the problem is. Great job!

[@MihaiDesigns]

Fingers crossed!

[@goury]

Only if using very soft filament very slowly, otherwise gears come apart and viola

[@ColinWatters]

Excellent detective work.

[@FlyingPanMan]

Just came across this video and the "woodgrain" effect made me chek through my discarded print box. Now I have a Mingda D2 with a direct drive extruder that is fitted with a single idler bearing, so not dual geared. Sure enough, I found a single skinned vase I printed in vase mode and, when held up to the light, it has the woodgrain. However, I suspect the bearing has very slight side play, but I'm not going to test it because a new dual geared extruder is being fitted tomorrow... and that's going to be used on flexible filaments.

[@Havalitoswe]

Amazing work man, I will surely check if this problem exists in my printer.

[@jooch_exe]

Excellent investigation, and now i know why i still have my Anet A8

[@Personnenenparle]

From your finding, i think you would need a pivot that its axis goes thrue the gear allowing the filament grip to move with a different stifness

[@Kaliumcyanidful]

Interesting finding! Well done! I also noticed this pattern on my prusamk3. But I did not notice it anymore since changing to 0.8mm nozzle 🤪

[@MihaiDesigns]

😅

[@PsychoShaft]

Just an idea, why not remove the teeth from the geared idler? so it free spins and set the tension on the spring screw to slightly tighter than normal? this would still be better than using a bearing as the place holder?

[@fredfred2303]

I found that the harmonic vibrations from my stepper motor/driver combination and rotational speed of the stepper can also generate patterns on the print, there are so many possibilities. The new large diameter planetary geared extruders with 10 to 1 or more ratios tend to eliminate this problem. For example, I have a single gear direct drive extruder (direct drive off the stepper motor with smooth idler) that creates patterns on the print, where my geared extruders 5.1:1 do not. Regards.

[@HelgeKeck]

This is the Gold Standard. Count me impressed.

[@olafschermann1592]

Nice surprise: i bought an Elegoo Neptune 2S with dual gear extruder. To prevent self-grinding i disassembled the extruder gear to install the shim fix. To my surprise they already have 2 copper washers installed. 👍

[@aecrim1969]

Awesome work Mihai! I love your tenacity to discover and solve small details like this. Keep up the good work! Hai noroc!

[@nobodynoone2500]

There is another issue in play you missed!! The excentricity, however small, in each gear, is always aligned the same, because the gears are the same pitch, varying extrusion pressure no matter the slack. Variations in tooth serrations are also an issue. In ME we would use 2 different sized gears, with an intermediate, but that would introduce more play. Time for a toothed belt! You can vary the size of one extrusion gear, and correspondingly to its drive gear section. That should solve both your observed issue and the one I see as well. Don't assume compromise is the solution. We have engineering.

[@MihaiDesigns]

Yes, was pointed out before. Eccentricity should be a very different pattern, way way more elongated. Tooth shape could be a thing as in the machining not being accurate enough. The play itself shouldn't be an issue so long as the motion can be transmitted uniformly, as we're only extruding one way. At least the way I understand things.

[@nobodynoone2500]

@@MihaiDesigns It might be tiny, but as we approach higher levels of precision, it's something to consider. Our craft is making new leaps recently.

[@NaterNorris]

I was thinking the same thing about the belts. A double belt extruder

[@drewporter8959]

@@NaterNorris there is one that has been around for a while, but is for Bowden systems. www.thingiverse.com/thing:2289203 Works really well for the printers that we use at work. A guy that started before I did switched all the Bowden extruders out for this instead.

[@christopherbray598]

I remember seeing this craze UM2 feeder years ago. www.youmagine.com/designs/ninjaflex-geared-ultimaker-2-feeder

[@DocMicha]

Now you made me think about this. Could it not be, that the teeth of the gears have to be synchronized to the teeth of the filament transport. The pattern you have looks in physics like a beat signal between 2 almost identical frequencies or multiples. It is: Tbeat=T1*T2/(2*abs(T2-T1)). Tbeat are your 1.5mm T1 and T2 are the two periods in competition.

[@MihaiDesigns]

I'm pretty sure you're right. Even though the larger T1 appears to have a higher amplitude. When it's gone, the other one is clearly visible.

[@Kyrazlan]

You sir, earned a sub. Seems to be a descepancy with the involute gear profiles not providing consistent torque through the gears meshing. The involute profile is designed to provide stable torque through the motion of the gear but when they are too loosely meshed you have create a situation where the motion is not transmitted smoothly. I wonder if helical gears would solve this issue.

[@MihaiDesigns]

I wonder the same. I might be getting such gears for testing in a while.

[@gordodefuego]

bro ive been dealing with for months now! thank you! edit: when I used TPU or any flexible filament the issues would disappear magically leading me to the extruder area but didnt know what to do about it

[@orphax1925]

engineering student here, you could use another indirect connexion between the two gears, for example you can invert the axial direction of one then connect the two with a set of gears that can allow the second to translate a bit depending on the filament

[@ParkerDD]

Or use timing belt (2 loops, one idler to transfer between the two) to drive it as well, no need to change axial direction of the gearing

[@fmphotooffice5513]

Excellent presentation. Logical, focused procedures.

[@MihaiDesigns]

Glad it helped!

[@jamespray]

Awesome work, thank you! I was not seeing the pattern with my LGX, and the biggest mechanical differences are that the gears are much larger than the ones you showed here, and the tension adjustment is quantized to set stations instead of spring-loaded. I'm trying to think of what difference those would make. Larger gears would mean that adjustments in the gear mesh would be smaller in proportion to the gear radii, so small variations might have less impact. The quantized tension adjustment seems like it wouldn't have much impact, except that it might reduce variation by holding the gear mesh to a set amount (versus the inherent flexibility of a spring-loaded arm)... Two other differences are that the large hob wheels have more surface area in contact with the filament, which might average out some variation, and the build/machining quality is hopefully better (given the price premium 😅).

[@Tedlasman]

the lgx is geared, mihai's isn't geared.

[@MihaiDesigns]

Yes, I was thinking the same. Will have to get hold of one.

[@jamespray]

@@Tedlasman Yes, the LGX is 3:1, but I'm not sure this would matter because Mihai's findings are pointing to the driven/idler gear mesh, which is 1:1 regardless of the stepper/output ratio. That drive geardown could certainly introduce artifacting of its own, though -- and that could be tested on any of the dual-drive geared extruders like the Orbiter, Sherpa, Hemera (I think?), M4, LGX, etc. These offer ratios from 3:1 to over 7:1, if I'm not mistaken, and would all produce different periodicity if it's a problem. I only have access to an LGX and a Nimble right now. My Nimble (early-gen, V1.2) does have hilarious levels of artifacting because of the drive cable, and my testing from that gives me confidence that the drive ratio won't be a primary factor -- the artifacting I saw repeated at about 1.4mm^3 of filament, which corresponded exactly to half a rotation of the drive cable (that is, upstream of the Nimble's 30:1 worm gear), and changing the ratio upstream of the drive cable with a 1:3 overdrive on the stepper didn't alter the pattern. So you have to consider how potential artifacting sources might stack, and in what order.

[@jamespray]

One other thing that might set the LGX apart is that the large gears should have more teeth interfacing at any given time, compared to the normal ones, which would also add an averaging effect. If you think about the mechanical problem, these small-geared dual drives push a fair volume of filament for every tooth interaction within the mesh, and I don't know if it's possible to have a straight-cut tooth profile that gives perfectly smooth motion through the full sweep of two teeth meeting and parting ways when the depth of mesh is variable. Any variation of where the teeth intersect will slightly change the 1:1 ratio one way or the other and introduce extrusion variations for straight-cut gears. On that basis, I'm convinced Mihai's suggestion of herringbone driver/idler gears would probably go a long way toward resolving this for gears of the size in the video.

[@Bordpie]

@@jamespray Standard involute gears do allow a change in the centre distance between the gears, while maintaining a constant rotation speed between the gears, this is one of the benefits of the involute profile. That's why this pattern effect is so perplexing, since in theory it should not happen. My current hypothesis is that it is caused by shaft angle misalignment, which is known to cause meshing errors and edge loading of the gear teeth (it would also explain why it happens on some extruders but not others). This is dealt with in industry by tooth crowning, which solves shaft misalignment issues but the only way to know would be to test.

[@BarsMonster]

Thanks for your findings! Last year I was fixing inconsistent extrusion and converted BMG to single gear by grinding teeth out. Single gear is ok for me as 99% of the time I work with hard filaments. With single gear you only need 1 high quality gear to rule the world... Will try your STL's for study.

[@ExplodingWaffle101]

just out of curiosity- are the remaining patterns vibration, like the ones fixed by input shaper?

[@MihaiDesigns]

On the blue model at the end? I doubt it. I've got another video on that (still to be edited), but I haven't tested my theory yet.

[@KeithSachs]

looks a lot like driver noise to me

[@RockIsLife001]

Input shaper is only used to prevent “ringing” look up the term on google. This is a completely different artifact. Input shaper cancels out ringing and pressure advance in combination makes edges sharper. These artifacts are not caused by either of these because they are moving in straight lines with no turns or changes to accel and decel.

[@NicksStuff]

Salmon skin, maybe?

[@oliviermiclette160]

I don't think the problem is due to the type of gears (spur gear). I think this problem comes from poor gear alignment and insufficient contact ratio to ensure a constant velocity. The involute profile of gears (if these gears are well manufactured) can allow some backlash while still keeping a constant velocity. But if the mounting distance of the gears increase too much (more backlash), the contact ratio will decrease below one (1), then you start having inconsistant velocity issues. The gears being not parallel to each other is, from my point of view, the biggest issue here. I don't think the dual gear extruder is a bad idea, it is just not designed properly. Thank you for sharing this content, I enjoyed your video!

[@AirsoftAbominations]

a possible solution: independantly tensioned gear, in that the spur side is either fixed or independantly tensioned to the hobbed side, you can then have the tilt angle vary on filament squish and thickness but stull have the spur teeth mesh perfectly

[@Bobis32]

i will say im not an engineer however if the idler bearing is slipping shouldn't you just increase the tension to the idler and add scoring since that will increase friction reducing slippage, this is just coming from someone who took engineering classes for 4 years in HS