Ask questions about or recommend improvements to Dr. D-Flo’s massive 4’x4’x4’ 3D printer build.
Are you caught up with the series? See the YouTube Playlist.
Also checkout the webpage for this build, which is packed full of supplemental information.
Summer brake is coming, I hear you will have a little bit more time for projects. What are you further plans for development and optimization of the large format 3D printer? Next models for printing?
Would you give a try to pellet extruder printerhead on the Elegoo Giga? It might need motor and structural upgrade too, but would have solved the expensive filament usage for large prints.
There are some cheaper less powerfull pellet extruders (than Massive Dimension) out there, like Mahor V4 and Pulsar Atom with respective flow rate of 200g/hr and 800 g/hr.
Such a printer still might fit on standard rated wall plug. I think the power line requirement of LF3DP is an additional obstacle for wider non industrial use.
I am thinking of modifiing the Elegoo Giga or making a pellet extruder printer from the scratch. Probably one of the requirements would be so the printer would fit through the door when assembled (I think that is a no go for the Orangestorm Giga). Assymetrical build of around 760x1200x1600 mm outside (fits through standard european doors) and 600x900x1300 mm inside so you can print a topologically optimized bike frame.
Hey Check 
,
Yes, I am spooling up for the summer! Lots of fun projects planned. The first part of the summer will be focused on finishing the upgrade to my large format printer (new MDPE10, 5 kg/hr extruder). The ETA for this video is about 2 weeks. I then plan to finally make good on a promise that I made nearly two years ago… 
Small pellet extruders (<1 kg/hr) are actually a brief topic in my upcoming video. To preface my following comment, I never want to stand in the way of innovation, and I am more than happy to be corrected, but the math doesn’t compute for small pellet extruders. For several reasons:
- They have a small feed throat, so incoming pellets are more likely to bridge and cause jams or inconsistent extrusion. Forget about irregular shaped recycled material.
- Their screws are far too short, which means that the pellets are receiving a majority of their heat from the heating elements in the walls of the barrel and not through shear/viscous heating (i.e., friction created by the screw). For this reason, you are still limited by thermal conduction, which is what constrains the volumetric flow rate of filament extruders. Further, because the screw is so short, you won’t be able to achieve adequate mixing of pellets and colorant/masterbatch, so you will have to buy pre-colored pellets if you don’t want your print to have color gradients. This will significantly increase the price of pellets.
- Even for small pellet extruders are heavy due to the hardened steel construction of the screw and barrel and the large motor and gearing needed to drive the screw. Printing infill has a lot of accelerations and decelerations and this added mass should not be ignored.
These observations are based on my experiences with the MDPH2 extruder, and is one of the reasons why I am so excited that the MDPE10’s screw is twice as long.
So for any applications under 1 kg/hr I would recommend the typhoon extruder. Using filament comes with the added benefit of being able to retract without artifacts. I understand that filament can cost almost an order of magnitude more than pellets, but having prints fail due to inconsistent pellet feeding or the post-processing needed to clean up stringing and blobs is also an important consideration.
1 Like
Row, row, row your boat …
Might be fun part of the video if you can ride a benchie down a river :).
Thank you for your invaluable insight. I have started 3D printing a month ago. I am on 3 pcs of Bambu P1S now, soon to add more capacity. I have transferred one of my most complex products from traditional manufacturing to 3D printing. After changing the models to get the most out of the technology I am very happy with the results, its better in every aspect - quality, time, price, labor.
I have other products I would like to transfer into 3D printing - these are much larger, heavier and not easily divisible for P1S buildspace. I am not just yet ready to commit to your LF3DP both due to financial and space reasons. Although as I have found out it is a well thougth out build and considering the performance on the cheap side. I am just not ready yet, hopefully soon.
Considering sticking a small pellet extruder on a consumer grade 3D printer I think you are just right. Questionable mixing with masterbatch pellets, low volumetric flow, low speed of printing due to heavy print head, belts and low torq motors. It just does not make sense even from financial point of view (low productivity of such a printer).
There is certainly a hype around the small pellet extruder projects. Claims about saving money on filament are easy to sell. Most of the projects are dead, some are just getting started, a few are in production, still a niche market confirming what was said above.
My models use fuzzy skin both as a final surface finish and to assist better glue adhesion. Therefore I need lots of acceleration to print fast. Also meaning I am stuck at 0,4 nozzle. Considering the models have mostly closed surface - there is not a lot of visible retractions - I am starting to think a bowden extruder is a way to go for my particular use case.
Some have replaced Creality K1 first version problematic extruder with a massive bowden one - shaving off 140 g of accelerated mass - getting twice the speed as a result. I would be very much happy with a machine capable of 50 mm^3/s or 180 g/hr.
Price of Typhoon extruder is too steep for me and does not solve the weight - acceleration problem.
What do you think about this aproach? Does the strong bowden extruder covers the middleground between large pellet extruders and small filament extruders?
I am thinking in terms of floor space vs. production capacity. I can surely keep adding the P1S, but there will be soon reached a limit of floor space. If I can add more capacity with the same footprint then it is a win-win. Real estate is of a premium here. This way I might get to the LF3DP faster down the roadmap.
Yes, the fact that a true small and affordable (<$1k) pellet extruder has not made it to the market is very telling.
I think it is clear to you what your needs are (limited floor space, high accels needed). A strong bowden setup seems to be suitable for your application 
1 Like
Upon further market research my hats off to you sir ! Very much looking forward to see next progress. Starting to think it might be a good project for 2025.
Regarding nonlinear extrusion speed dependency on auger rpms. Did you overcome it already or are printing on constant extrusion and constant speed? I know slicers can not work with it.
If it is a specified or measurable constant nonlinearity and can be calibrated into a single nonlinear curve, this can be applied directly to the G-Code via simple text search and change operation. In PHP it would be probably one line Regex expression to change the extrusion value over speed. Nonlinearity would be applied from single Lookup table of values (tabled nonlinearity), perhaps with some interpolation. It does not have to be even a math equation. So it could be implemented as a webbased service. Seems to me straightforward sollution. Maybe you have already done it. Anybody with PHP or JavaScript website knowledge can implement it once the nonlinear function is calibrated. I am very bad at Regex expresions - it is not my daily bread and butter. But somebody else can be helpfull - it can be done fairly easy.
Hi Check,
RepRapFirmware has correction for nonlinear extrusion similar to how you explained. The g-code for it is M592 and here is a link to the docs.
While I have found for there to be some nonlinearity in screw extrusion, the bigger issue is the time it takes to build backpressure in the barrel between printing moves. For filament 3D printers, this hysteresis or lag is corrected with pressure advance. However, even small values of pressure advance can trigger a “retraction.” In screw extrusion, spinning the screw in the reverse is a big no-no. In both my MDPH2 and MDPE10 extruders, the screw has some vertical play in the barrel. This is not an issue under normal operation as the backpressure pushes the screw all the way up. However, “retraction” will cause the screw to drop. Also, “retraction” can pull melted material back into lower compression screw flights which can plug the screw.
I have found two solutions:
- Maximize the length of print moves by:
- Enabling Arc fitting (G2/G3) to prevent curves from being broken up into small lines
- Lowering the slicer’s resolution as to not generate too many segments
- Ensuring that the extruder is not changing flow rates and widths when printing a feature. This is especially important for a uniform and aesthetically pleasing exterior perimeter.
- Print really large! Large extrusion widths are less impacted by this back pressure problem. Also, small extrusion errors are less noticeable on large objects.
1 Like
Nice, one problem solved already, another two raised. So the generall rule here, as some of the problems you have mentioned are currently unsolvable, is to design the model according the specific needs of the pellet extruder. Which is a good rule of thumb in any field anyways. And large nozzle sizes dictates that layer lines are visually part of the design.
This is an obvious, yet a new way of thinking about 3d printing for me. It really shows how much of a newbie to 3D printing and CNC I am :).
Is there some play with the reverse operation of the auger as to perhaps:
1/ allow some retraction, counting how much, stoping it at some point
2/ lowering some of the heat zones temps to prevent heat/melt creep
3/ making all the above part of hysteresis curve meaning allowing small parts to be retracted/reversed and then waiting to clean the auger in normal mode of operation before retracting again.
Reading up on M592 it is only using a single quadratic coeficient, no hysteresis yet.
To continue my back of an envelope analysis.
Brute force hysteresis calibration on a pellet printer is insane. One would need 9 different speeds of extruder to optimally use the speed and part cooling ability of printer. Then one would need to calibrate nonlinear transition from 9 states to the other 8 states meaning 72 combinations of extrusion transients and 9 static extrusion parameters. All this for every material you want to print, maybe even for every supplier, maybe even for every color. Insane…
Somebody needs to implement this paper in the world of pellet 3d printing:
Would you be interrested in me connecting you with someone at Czech Technical University in Prague? I think it would be a great subject for a thesis. If you could describe the problem in detail, somebody can comeup with model solution in Ansys, Comsol or Matlab and you could test it.
Hi Check,
You will need an extruder that mechanically retrains the screw (e.g., tapping the tang of the screw) and then you could achieve something as seen in this video. However, I have discussed this seemingly impossible retraction with Sean Miller from Massive Dimension and we both agree that in addition to the screw being mechanically retained, this was achieved through:
- Using a very low extrusion temperature, “which would make it easier to pull back, instead of the polymer just stretching” (sean’s words). But low temperature would lead to poor layer adhesion.
- Using a small diameter nozzle (<1 mm). Again, if you need detail and plan to print with small nozzles opt for a filament extruder (my opinion)
This is complex. Yes, you can lower the the temperature of the metering zone (the one closest to the die/nozzle) to decrease the viscosity of the extrudate to keep it from oozing. But in practice I have found that you have to lower the metering temperature to a point where you get poor adhesion with the next layer. Also, you can only remove so much heat through the barrel. Changing the temperature of a heating zone really only affects the material near the surface.
A hysteresis curve would be very complicated to calculate if not impossible to perform in real time. Unlike filament extrusion, most of the heat used to melt the material comes from viscous heating (friction generated by the turning of the screw). The temperature profile of the melted regions in the screw extruder would change with the velocity of the screw, which is constantly changing due to the accels and deaccels of the print head.
I believe that the most practical solution is to put the extruder on a gantry or robotic system that can move the pellet head so quickly that oozing is not an issue. This is the reason I upgraded to Rack & Pinion in my last video.
Unfortunately, calibrating the extruder is my least favorite part of this project. I strongly prefer the machine building and CAD side. So I am going to stick with optimizing toolpaths and increasing travel speeds between printing moves versus complex modeling of the rheological behavior of the polymer melt.
1 Like
I like the aproch - if you can not fix it - go faster 
.
Understood and agreed. Difficult to navigate complex project to keep the upgrade pace steady while having a working machine at the same time. One has to choose his battles.
I will try midsize printing with Rat Rig Core 4 500 mm - I am on the waitlist. See where it leads. There will may be an opening for LF3DP in 2025 if I figure out a place for it in the new house.
Seeing how much time end effort you have put in to avoid rabit holes in the project, it is the only reasonably option right now. Starting to see a point in buying somewhat ready made solution with some sort of support is better for me, even if it is absolutely speaking expensive, but relatively cheap. In a years worth of printing it pays off.
Please consider LF3DP for dummies as a part of next video.
Hi Dr. D-Flo,
I hope you don’t mind this sort of brainstorming.
After watching this video:
Solution to adding detail to LF3DP (any large printer) is kind of obvious. IDEX printer. Outer detailed skin made with filament extruder of 1X size (or wherever needed) inner structure made with pellet extruder of 3X (or other) size. Taking the best of both worlds and IDEX is already inside RepRap firmware.
Oozing of pellet extruder can be solved with a silicon brush (Bambu inspired). If there was a separate carriage for filament extruder it can brush off the nozzle of pellet extruder with soft compliance bistable mechanism activated brush (switching it on and off by bumping the filament carriage to the frame). Bambu machines use the bumping to the frame to cut the filament when changing it with AMS. Filament cutter is only a razor operated by lever and this lever is bumped to a spring loaded finger on the left, inward, top side corner. It is a small nonprintable volume setup in the slicer. If all printable volume is needed, the finger can be diverted off and slicer setting changed to print everywhere.
Thank you for the extrusion article update and the new video. Both are awesome and answered a lot of my questions.
Cause of unstoppable tinkering with it, I have discovered a project for optical measurement of filament diameter and its online extrusion compensation via feedback loop in modified Marlin firmware.
I find it interresting in relation to the pellet extrusion hysteresis problem. If using an optical sensor to detect extrusion width on buildplate one could setup an automatic calibration procedure at the start of each print - shifting speeds - measuring width and compensating for it automatically. In similar fashion like the Bambu X1C does. Or this guy made an open source calibration device with camera and laser:
I think the solution is within reach. It is on the roadmap of integration this solution in Klipper.
To continue my serenade I went the deep pellet extruder rabbit hole even more and “discovered” Harmonic Drive gears and motors.
Pretty clever design getting huge gear reduction and torque in a small and light weight package. Miniscule wiggle. I think ideal for pellet extruder application. It is expensive, but the significantly lowered weight of the printhead and therefore much fast possible print speed is worth it in my mind.
