extruder nozzle design - why they are bad

Hi,
I'm new to 3d printing. I'm in the process of researching my first printer, how they work, the capabilities and limitations of 3d printers. I do, however have 12 years of manufacturing experience with plastic extrusion and injection moulding. Maybe this has given me and insight or maybe this has been discussed before?

The main problem as I see it with 3d priting is the visible ridges of each layer. They mar what would otherwise be a nice print, or create lots of extra work in finishing up the print properly.

The circular shape of the extruder nozzle is responsible for the ridges. It lays down a squashed oval cross section track which leaves a visible ridge.

Has anyone tried a square or rectangulare nozzle? Instead of a squashed oval cross section, the track would be a squashed rectangular cross section. Ridges would be less prominent.

I realise this would work if the nozzle is moving along the axis of the square nozzle and create problems in any other direction - if the nozzle was going diagnally, the cross section would be a squashed diamond. The ridges would be even more prominent than normal.

A solution would be a rotating nozzle that aligns along the path of the extruder. Obviously, engineering this and programming the software for this is well beyond my capability.

Has this solution been thought of before? Is it worth a try?

I can provide pictures of my ideas if requred.

Martin

You can't easily drill a small square hole. However, you can do this with a laminated constructions. If you are going that far and you are going to add another degree of freedom, why not actuate the nozzle opening. This will allow you to use a small square hole for detail work and a large rectangular hole for fill work. As you have already alluded to, this requires a completely new slicer program.

I like this because a machine like this could also be used for a pick&place, cutter, etc.

Edit: On second thought even if a square nozzle works as you propose it does, it will only make the vertical walls better. It would make the ridges on inclined surfaces more pronounced. I still like the idea of a variable nozzle.

Edited 1 time(s). Last edit at 05/15/2014 06:50PM by nicholas.seward.

---

ConceptFORGE
Wally, GUS Simpson, LISA Simpson, THOR Simpson, Sextupteron, CoreXZ

My guess is the strength and accuracy of the actual machine is more at fault but it could be the electronics I haven't run a machine with the standard reprap electronics yet. I did build my own very rigid floorstanding 3-D printer and the results are fantastic but I did use CNC control software. I bought a Arduino ramps 1.4 set up with drivers for $59 and I'm building a small printer now to see if it is the electronics at fault here is a picture of a bottle opener I made on the large machine if you look closely you'll see how accurately the beads are laid down

Quote
martinfromoz
Hi,
I'm new to 3d printing. I'm in the process of researching my first printer, how they work, the capabilities and limitations of 3d printers. I do, however have 12 years of manufacturing experience with plastic extrusion and injection moulding. Maybe this has given me and insight or maybe this has been discussed before?

The FDM/FFF type of extrusion printing has been around for the better part of 30 years. Checking out some of the patents particularly from Chuck Hull will give you insight to what's been going on. It's not the shape of the orifice in and of itself as there many other issues at play. My suggestion would be to build a starter printer and get familiar with the process. There are some similarities with other processes but it's a process in and of itself and has specific limitations and parameters that aren't either obvious at first or behave in a way that someone without the background in FDM/FFF would be aware. Specific to layer ridges there are some other issues to consider. By the time the next layer is laid down the previous layer is already cooling and add extrusion control and whatever mechanical abnormalities the machine introduces and it's not so cut and dry. For example there is a phenomenon known as "Z wobble" that is responsible for many if not most of the layer irregularities people encounter. There will be slight ridges on even the most sophisticated and advanced machines though certainly not to the extent of the less capable machines or machines that don't use quality parts or accurate assembly and calibration.

With patents expiring and the current round of SLS and SLA printers coming online there will be some opportunity to address some of the issues of the FDM/FFF process. I think once you get in and get some hands on it will give you a better perspective to base your decisions on what can be improved. No doubt extrusion is an area for improvement but it's not going to be as simple as changing the geometry of the nozzle outlet.

Quote
vegasloki

Quote
martinfromoz
Hi,
I'm new to 3d printing. I'm in the process of researching my first printer, how they work, the capabilities and limitations of 3d printers. I do, however have 12 years of manufacturing experience with plastic extrusion and injection moulding. Maybe this has given me and insight or maybe this has been discussed before?

The FDM/FFF type of extrusion printing has been around for the better part of 30 years. Checking out some of the patents particularly from Chuck Hull will give you insight to what's been going on. It's not the shape of the orifice in and of itself as there many other issues at play. My suggestion would be to build a starter printer and get familiar with the process. There are some similarities with other processes but it's a process in and of itself and has specific limitations and parameters that aren't either obvious at first or behave in a way that someone without the background in FDM/FFF would be aware. Specific to layer ridges there are some other issues to consider. By the time the next layer is laid down the previous layer is already cooling and add extrusion control and whatever mechanical abnormalities the machine introduces and it's not so cut and dry. For example there is a phenomenon known as "Z wobble" that is responsible for many if not most of the layer irregularities people encounter. There will be slight ridges on even the most sophisticated and advanced machines though certainly not to the extent of the less capable machines or machines that don't use quality parts or accurate assembly and calibration.

With patents expiring and the current round of SLS and SLA printers coming online there will be some opportunity to address some of the issues of the FDM/FFF process. I think once you get in and get some hands on it will give you a better perspective to base your decisions on what can be improved. No doubt extrusion is an area for improvement but it's not going to be as simple as changing the geometry of the nozzle outlet.

Z wobble can be eliminated by using a ball bearing spline. Like Leapfrog printers use.

There are more cost effective ways to eliminate Z wobble than with ball screws. A good ball screw like those Thomson parts command a premium price though they are top quality kit.

Here is a thread where a Tantillus user used the standard threaded rod for a lead screw and then moved onto a ball screw and then finally to get rid of the wobble he went with the cable driven Z axis [forums.reprap.org] . Lead screws regardless of the type are not the best for a 3D printers. 3D printers are best made using engineered solutions rather than brute force solutions used for CNC. That is not to say an overbuilt CNC style machine is bad just that you can get just as good of results using engineering to overcome the need for a 1000lb machine.

This thread shows one of the most intricate and large prints I have ever done and it used a cable Z axis machine. [forums.reprap.org]

---

FFF Settings Calculator	Gcode post processors	Geometric Object Deposition Tool Blog
Tantillus.org	Mini Printable Lathe	How NOT to install a Pololu driver

In my opinion it's the strength of the guides that's the problem 8mm or 10 mm unsupported round bar is very easily flexed

There are a hole host of issues at play, 3D printers are collections of compromises, they are in general not very stiff, the bearings are cheap and tend to have significant slop, but they are in general good enough for what they do.
I fought Z Wobble on my MendelMax for months, tried pretty much every suggestion, my final fix was to move to Bronze bushings on the Axis.

I have seen really good prints from lead screw machines, but I will say that Delta printers tend to have spectacularly good layer alignment, as does the PrintXel which uses a belt driven Z axis, and my current belief is that lead screw based Z axis are just not a good match for most 3D printer designs.

---

___________________________________________________________________________

My blog [3dprinterhell.blogspot.com]

Quote
Polygonhell
I fought Z Wobble on my MendelMax for months...my current belief is that lead screw based Z axis are just not a good match for most 3D printer designs.

Leadscrews are perfectly fine for 3d printing. I use them on my printer and I have zero banding. The problem with the MendelMax isn't that it uses leadscrews. Most 3d printers are more that adequately "stiff" for what they need to do. Tons of printers use leadscrews for the z-axis, and I guarantee 100% that if you had a printer that was easier to align, your problems would have been solved months ago. The problem with the MendelMax is poor adjustability and a overly complicated frame geometry, which makes alignment extremely difficult and in some cases impossible.

I own a MendelMax that I haven't used in over a year. I had a b**** of a time getting the leadscrews aligned well enough that I felt the print quality was acceptable. It has precision leadscrews. I have another printer that I built. It uses some cheap-o ACME leadscrews (Imperial threaded even! OMG the worst crime of all!) from the hardware store. The bottoms of the screws are unconstrained and do not wobble even a little bit. I see zero artifacts from the leadscrews. Why? different design that's dirt simple to get the screws to align.

Quote
Sublime
That is not to say an overbuilt CNC style machine is bad just that you can get just as good of results using engineering to overcome the need for a 1000lb machine.

A thousand times yes. Nail on the head, Sublime.

Rigidity or "stiffness" are generally unquantified terms people throw around when they don't know how to engineer a machine for its intended purpose. The belief that throwing more and more money at a printer and buying more and more expensive parts will yield better prints is a fallacy that has been proven wrong with 3d printers many times over. The fact that you can stand on your printer does not mean that it will print better... it only proves that you don't understand the physics 101 concept that force is a vector and the directionality of that vector might just be perpendicular to the axis where you could actually benefit from it.

I guess it all depends on what you think is acceptable but especially with the extruder is mounted to the X carriage reprap designs are not strong enough on the z-axis in my opinion. If You run the figures on the guide rails of the Z with decent acceleration rates on the x-axis the amount of flex is unacceptable to me anyway. But like you said trying to put something together that has a strong frame with rails which are continuously supported is a very tough job and time-consuming on un machined surfaces.

Edited 1 time(s). Last edit at 05/16/2014 09:35PM by cnc dick.

I'm sure it's possible to have a heavy enough extruder with high enough accelerations to flex the leadscrews. But even within that statement, there are many variables to consider. What are the diameter of the rods? How long are the rods between the constraining points? Carbon steel, plain steel, stainless steel? Are the rods constrained in the x-direction by both x-ends or just one?

I believe that in most cases with hobbyist-designed 3d printers (where very little modeling is ever done in the design phase), good implementation of "standard" parts will yield equal or better results than poor implementations of expensive parts. You don't always get what you pay for, and expensive things touted as upgrades might simply be a patch to add needless "rigidity" to what is essentially a design flaw.

No design is ever perfect. I just think that if you can solve a problem with the parts you already have and the parts that people generally use, then you have contributed something to the collective knowledge. If you just go buy a bunch of expensive linear slides, ballscrews, aluminum extrusions, etc then you haven't really done any engineering. You've just substituted money for knowledge.

Edited 2 time(s). Last edit at 05/16/2014 10:18PM by iquizzle.

Quote
iquizzle
I'm sure it's possible to have a heavy enough extruder with high enough accelerations to flex the leadscrews. But even within that statement, there are many variables to consider. What are the diameter of the rods? How long are the rods between the constraining points? Carbon steel, plain steel, stainless steel? Are the rods constrained in the x-direction by both x-ends or just one?

I believe that in most cases with hobbyist-designed 3d printers (where very little modeling is ever done in the design phase), good implementation of "standard" parts will yield equal or better results than poor implementations of expensive parts. You don't always get what you pay for, and expensive things touted as upgrades might simply be a patch to add needless "rigidity" to what is essentially a design flaw.

No design is ever perfect. I just think that if you can solve a problem with the parts you already have and the parts that people generally use, then you have contributed something to the collective knowledge. If you just go buy a bunch of expensive linear slides, ballscrews, aluminum extrusions, etc then you haven't really done any engineering. You've just substituted money for knowledge.

I agree that just throwing money at it although it may work probably isn't the best design. But I've designed plenty of machines in my life and if you run the figures for the vertical z rods with a extruder mounted to the X carriage the flex in the center of 14 inch piece of round Rod in the middle is terrible it is not designed correctly. Why everybody keeps bringing up the screws always baffles me it's the guide rods that are at fault this is what you design to hold the axis and this is where Mendel reprap is wrong in the design. That's why people like light designs Bowden extruders Delta style printers all of this is because the z-axis are not stiff enough obviously the Bowden lightens the mass in the Delta redirects a lot of the moving mass to up-and-down and also has a very light X and Y moving mass

Edited 2 time(s). Last edit at 05/16/2014 10:42PM by cnc dick.

Thanks for the carefully considered replies.

Nicholas.seward, It would be interesting to try. A square or rectangular nozzle would have to be made by constructing it in several pieces as you mention. I saw a video of a concrete printer contour crafting . The printer had a 3 in one extruder. Two nozzles for the perimeter of a wall and a larger moving one for the fill. At the start of the video is a clay printer with a rotating nozzle. it has very smooth exterior walls. So I guess it has been done already.

Do you think a better designed machine would be less likely to cause banding? Better calibration?

I intend on building or buying a delta printer when some funds come in. Hopefully I get to experience all this for myself.

@martinfromoz: Build any printer. It doesn't matter. Don't worry about banding. Start printing and then you can make tweaks to fix any problems that you find. It can be debilitating if you worry about doing it the "best" way when you don't have the experience to really judge. Just get in there and start printing.

Look in my signature for a list of "bad ideas". Some really are bad ideas but they all have added to my experience. I like to ditch conventional wisdom and rely on calculations and experimentation. Nothing will ever change if you rely heavily on conventional wisdom. Keep thinking crazy thoughts.

---

ConceptFORGE
Wally, GUS Simpson, LISA Simpson, THOR Simpson, Sextupteron, CoreXZ

relevant (see idea 9):

http://makezine.com/2013/05/02/first-to-file-nah-first-to-blog/

To some degree, there is a point where you over-engineer the solution. In the obvious respect, most printers at this level are about attaining some degree of accuracy of the commercial units without spending the cost. I could give many examples, for instance the always common Z drive mentioned earlier. From a certain point of view, people have mentioned the $3 threaded rod, the $10 ACME screw, the $50 ball screw, and the nearly costless fishing line drive. Depending on the application, all of them are adequate. Also know that what people want drives these solutions. Some people want accuracy, others want speed, and some are purely about aesthetics.

But going by the original idea, why not go a step further and make the nozzle rectangular such that it lays down like a strip of tape. There would be quite a few advantages to this, including the reduction of the number of passes to lay down the material. But I would have to come back to the effort vs payback on this. You would have to program a swiveling head as perimeters (the only place you see the edges) can vary in shape on more artistic models, let alone curved geometric shapes. This swivel would have to be dead-center, otherwise you would create a similar effect that backlash on X-Y or banding on Z would make. And then there is not just the programming. Making that head swivel is not adding a sub-routine to the firmware. Its adding the support to the slicer program because the hardware only executes gcode, the slicer would need to create that gcode.

So by the time your done, we have added an extra axis that would have its own development problems to work out (CNC has been around a long time, yet reprap is still trying to develop solutions for basic cartesion movement). You would also have to machine the parts for this new nozzle along with the swiveling head. The nozzle would be beyond basic shop tools to manufacture. Then you need to add another custom axis to your firmware, probably the easiest part of this entire design. And finally get someone to make a custom slicer.

And in the end, theoretically you would see what difference? If you see the output of a well calibrated printer built with a solid frame printing at 0.1mm layers, you hardly see the differences in the edges at all. READ: You see the edges, you just dont see the differences (or variances) in those edges. And your idea only addresses the variance in the edges, not eliminating the edges themselves. This is achieved by laying down a 0.1mm layer from a 0.4mm nozzle, and pretty much squishing out a more reliable amount of meterial to have a 0.48mm width. So many factors play into the reliability of this particular situation including: 1. Extruder torque being constant, since if the extruder 'pulses' it will lay down a hour-glass shaped thread. 2. X-Y backlash, as the edges will over-lap in bad places if the X-Y axis cannot maintain accuracy to much better than 0.48mm. 3. Z banding since obviously the amount squished to the side is proportional to the amount extruded and the height of the layer.

The point I am making is that there are quite a few places for improvement. But the focus should be based upon 2 things: Effort vs payback and also Need. Right now, there is very little need to improve the edges compared to what is already been done. And the effort required to improve this particular problem is way beyond most people ability, and the cost for a purchase solution would not justify the real-world results achieved AFTER all the kinks are worked out. Instead, there are many NEEDS yet to be addresses properly that have greater potential. Increased speed. Multiple materials or colors. Extrusion that has more tolerance for 'bad' filament. Mixing materials during extrusion.

With your background, you could provide invaluable insight into home-made filament extrusion. Take a look at Lymans filament extruder and see what I mean. The guy took a $15 auger and $3 pipe and made a basic extruder. Maybe you could come up with some ideas to improve that process. It currently extrudes a roll of filament at 8 hours a roll. Everybody would love it if there was some improvements to extrude at 1 hour a roll.

---

"Never argue with stupid people, they will drag you down to their level and then beat you with experience."

Rotates about the center line of the hot end:

Additive wire laying
[forums.reprap.org]
[forums.reprap.org]