Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
Hi everyone !
I 3dprint since 2 years now and i consider those 3dprinter as machine tool.
I began with a CR10 and now, tired of bed leveling, repetability issues...i have a MK3s and i know the next will be home made.
So i make usefull parts and in particular parts for automotive and more particular, parts for VW T3 (or vanagon for western people).
The parts i'm making need to be near injection molded quality and it means all of the parts are plain material.
What do we need to print the best quality parts with FDM ?
- A good mechanical accuracy of the machine to extrude where it's needed (MK3S is excellent).
- A precise control of the material flow to extrude what it's needed.
- A good layer before the one which is extruded.
First layer is on the bed and it is what it is, but the other layers needs to be on a good layer and for my concern with plain parts a good layer is a layer with mirror surface finish.
I suffered many troubles printing my parts and the majority of them are flow related, bad flow can ruin your surface finish and defaults are migrating on the next layer.
I have to say that i use a tricky filament, Ninjatek Armadillo.
By printing external perimeter first, last line of extrusion is held between two previously extruded line and can be compressed if your flow is too high and causing a bump in the surface and mechanical property of the parts are bad if your flow is too low.
After many hours observing my prints, my mistakes and my tries to resolve them, i went to a theory :
The flow is nozzle geometry and fluid pressure related.
For the pressure, linear advance can tweak this (I have to say that i use a firmware with linear correction 1.5).
The point is "nozzle geometry related".
All the slicers take a theorical nozzle diameter in consideration and the flow correction is filament related.
I think this is a mistake.
We need to know about the real nozzle diameter and adjust the pressure in consequence to obtain the best surface finish for the next layer.
Next mistake is how slicers play with dimensionnal accuracy.
It's said that for dimensionnal accuracy you need to tweak flow correction of the filament, but i think it's not true.
The path of the extruder is width of extrusion related, not flow related, so, for dimensionnal accurate best quality parts we have to find first the possible best width for the possible best quality layer and correct dimensions by tweaking the 3d model.
As i said "pressure is tweaked by linear advance", for the best possible layer, the quality of the deposition of every single line needed to make this layer is nozzle geometry dependant.
For a given nozzle diameter, you can extrude a precise width with the best quality, it's said that for 0.4mm nozzle it's 0.48mm, but when nozzle diameter is 0.421mm instead of 0.4mm the ideal width become more 0.50mm than 0.48mm.
If you take E3D blueprint, a V6 nozzle has a tolerance of +/-0.05mm, so ideal width can vary in some proportions.
To overcome the knowing of the real nozzle diameter, i made somme tests :
Let's take a relatively not pressure dependant filament like PLA.
By printing single wall cylinders and study the result quality, you can precisely adjust the ideal width of extrusion.
Next, by printing plain cyliders or rings with at least 8 outside and inside perimeters and study the surface finish after some layers ( to let the machine compensate the first layer potential over extrusion), i can look for my mirror surface finish by tweaking the nozzle diameter and correcting the flow by this method instead of tweaking it by the filament flow compensation.
By this method, i found that 0.417mm gave the best quality surface finish even with rectilinear infill (with 0% perimeter crossover).
0.417mm for a 0.40mm nozzle, that's not a crazy difference, but from what i have observed very effective in terms of quality.
Another trouble for my concern about the mirror surface finish of every layer is external perimeters.
The width of the tip of a 0.40 E3D nozzle is 1mm.
When your printer extrude a line, the tip of the nozzle "iron" the half of the previous line creating a little lip on the second half of this line, for an external perimeter, this lip will stay until the next layer.
When you have a vertical wall it's not so important but for overhangs, i think it is.
I think a good practice for the slicer would be to program an "ironing" pass on the exteral perimeters before the next layer.
And at least for now, i think that Prusa slicer need to implement an excellent fonction that can be found in Kisslicer, alternance of perimeters and infill penetration (decimal number of perimeters in Kissslicer).
Maybe it's a little bit long or maybe confuse by bad english mistakes (french is my native language) but i tried my best.
What do you think about all this ?
RE: Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
By printing single wall cylinders and study the result quality,
A cube allows for better measurement instrumentation accuracy. And is what most of us use. A cube also eliminates mechanical step accuracy from impacting wall thickness.
alternance of perimeters and infill penetration
P.Slice already has perimeter width and infill overlap settings.
RE: Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
I think a cube is not the best shape to take measurements, corners are always fucked, but no matter, you can use a cube if you want.
But i was not talking about measurements, i'm looking for a method where you can find the best printing parameters only by studying hoizontal, vertical or whatever surface finish and non sense about some parameters like flow correction which is filament related...
In Kissslicer, when you parameter by sample 2.5 perimeters, you have 2 layers with 2 perimeters, and one with 3, so infill edge is taken between 2 layers of perimeters, is it more clear ?
I think it's also a nonsense to make an overlap between perimeters ans infill giving a percentage, in fact you create a zone of overextrusion and so a default on the layer.
RE: Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
When you print parts with less than 100% infill, precise flow control is not a real concern, the machine can "push" defaults of extrusion inside the infill.
By defaults of extrusion, i don't mean patterns on the outside, i mean real overextrusion that can deform your part, create backpressure on the tip of the nozzle or under extrusion and all of it's consequences..
When you print at 100% infill, flow must be precisely determined, overextrusion gives bad surface finish (scratches on the linear infill...) and those defaults amplify at every layer.
Like towers for temperature or linearity correction or whatever test, printing a ring (by sample 40mm exterior diameter and 20 inside by 10mm high) with 100% infill is a tricky part if the flow is not precise.
Surface finish observation of the layers of this ring is the best way to adjust the flow, better than measuring a single wall of 0.45mm.
My point is i tweak the flow by modifiying the nozzle diameter instead of the filament related flow correction, bigger diameter if it need less flow and vis versa.
I think it's more logical.
RE: Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
The question is always what do you want to achieve? Based on my experience FDM printing is never 100% accurate. You can tweak it with all the given parameters in the slicer but it will only work for current filament and the current nozzle. Even ambient temperature and speed can affect your accuracy. And then you're printing different geometry and realize the settings does not apply to all shapes and of course not to all complex shapes. Then you have different plastic which shrinks in different way.
Once you switch filament or your nozzle get older, more used or need to be switched then you end up at square one again.
So far the defaults PR settings are producing pretty good quality for most of the stuff I'm printing. If I would need really high precision, I would go with a different technology and not FDM. Otherwise you end up more tweaking then printing.
Often linked posts:
Going small with MMU2
Real Multi Material
My prints on Instagram
RE: Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
I believe the OP is going to design and produce his own printer. Along with that printer, he will also create a slicer that accommodates all his preferences and findings. I just hope he includes the effects of micro-stepping and computational inaccuracies of print vectors not aligned with the three proper axis. I suspect his printer design will have to rely on servos rather than steppers though. Steppers are much too persnickety to control in a true linear fashion. Each step of an extruder an impulse change rather than a true controlled linear pressure change.
I am especially hopeful for a solution to the skin effect and surface tension correction to get filament to make true square corners. I hate the radius left on my square parts.
RE: Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
Ouch ! No need to be sarcastic tim-m30, i'm not agressive, i'm just trying to figure why things are like this and not like that.
You're right nikolai-r, nothing is 100% accurate and i'm aware of that.
I'm just a tech, sometimes more vulcan than Spock and i try to find the logic behind things.
So let's try again :
Nozzle geometry is a parameter of flow calculation, why do we use a generic diameter in slicers instead of having a method to determine it ?
RE: Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
Maybe you can make your issue clearer to more of us by showing before/after pictures of your optimization and what look you are going for in your parts.
I guess I somewhat understood what you are trying to do but to my understanding you just make the slicer lay down the extrusions closer or further apart from each other at a given flow thus reducing "overlap"? Wouldn't you have the same effect by simply tuning extrusion multiplier? To test that you would probably have to print the same exact GCode with two nozzles of known sligtly different bore?
RE: Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
[...] The point is "nozzle geometry related".
All the slicers take a theorical nozzle diameter in consideration and the flow correction is filament related.
I think this is a mistake.
We need to know about the real nozzle diameter and adjust the pressure in consequence to obtain the best surface finish for the next layer.
An interesting thought, but exactly how can you know the precise size of the nozzle opening? The slicer operates independently of the printer. There is no real-time feedback. The same gcode can be carried to many printers, or the nozzle changed between prints of the same gcode file.
The printer provides no real-time data to monitor of any sort. It just does what it's told by the gcode.
Given those considerations, I'm not sure how you'd propose fine-tuning the slicer output to the exact nozzle opening diameter at any time. Am I missing something?
And at least for now, i think that Prusa slicer need to implement an excellent fonction that can be found in Kisslicer, alternance of perimeters and infill penetration (decimal number of perimeters in Kissslicer).
It's an interesting feature that I believe Cura also provides. The idea is that the alternating number of perimeters locks the infill into the outer perimeters to provide better vertical support. I can't say that I've seen any analysis of whether this actually does anything or not, and I do wonder just how much of an issue "shifting infill" is in reality. Every perimeter adds to print time, and the consensus seems to be that infill is not as important as perimeter count for actual part strength, so this seems (to me at least) to be an unhappy medium that doesn't quite offer the advantages as other methods. Testing would be interesting.
and miscellaneous other tech projects
He is intelligent, but not experienced. His pattern indicates two dimensional thinking. -- Spock in Star Trek: The Wrath of Khan
RE: Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
Sarcasm aside: let's take a look at a few assumptions you make:
Let's take a relatively not pressure dependant filament like PLA.
PLA is VERY pressure dependent: when it melts it is a FLUID and becomes susceptable to hydrodynamic forces..
By printing single wall cylinders and study the result quality, you can precisely adjust the ideal width of extrusion.
Which is why I said measuring a flat is much preferable to measuring a cylinder to maintain experimental accuracy.
Next, by printing plain cyliders or rings with at least 8 outside and inside perimeters and study the surface finish after some layers ( to let the machine compensate the first layer potential over extrusion), i can look for my mirror surface finish by tweaking the nozzle diameter and correcting the flow by this method instead of tweaking it by the filament flow compensation.
You are ignoring the fact the bed is NOT stationary and can move underneath the nozzle. The frame and mechanics change dimension as they change temperature. It is unavoidable in the printer design. And unless you somehow map this displacement with extreme accuracy changes in layer dimensions can't be avoided and will nullify such testing/tuning. If the bed warps upwards 0.02 mm, your extrusion test is invalid (20% error with 0.1 layers). And the bed warps that much even with 60c temperatures; and much more when the bed is heated to 100c.
RE: Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
timo-m, my point is not my specific problem (or at least, i think so...), i'm not the only one who wants to print accurate plain parts (or am I ?).
bobstro, i've read your blog and found very useful informations, thank you 😉
There are multiple use of the 3D printer, some of us use them to make trinkets or toys or whatever which has only an esthetic function where approximation is not a problem..
For me it's a machine tool and i (or i would to) make functionnal parts, so full control of the fabrication process is needed.
For now, i can't figure out why the nozzle diameter which is an important parameter to flow calculation is an approximation.
In the world of machining, if you want accurate milling or turning operations, you have to proceed with an accurate measurement of your tools, the machine only execute a path (like our 3d printer) calculated by a software similar as our slicers but it takes account of real radius or real length of tools, not a generic value like 8mm endmill.
I think there's a nonsense here, we use a machine tool (a 3d printer) and as a base of material deposition flow calculations, we use an approximation, how can the parts be somehow accurate.
My intuition (and it's only that for now) is that there must be a method to be more accurate on the flow calculation by tweaking the nozzle diameter in the slicer to a less approximative value and it will be useful for every part and every filament.
I think that with trials and error of printing 100% infill parts and study the surface finish it is possible to determine a accurate approximation of the nozzle diameter which can be useful for all parts we make or all filaments we use, even if there's left some minor tweaks to do.
RE: Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
i think the issue is that the nozzle is not the active output of the process but only an intermediate tool to the real output ( the extruded bead of filament). in a subtractive machine tool the tool bit is the final interface to the work-piece and even there you need to compensate for flex of the tool and temperature compensation of the machine. with FDM you are extruding a viscous stream of plastic that will move relative the the nozzle as the nozzle changes direction. for good adhesion you want to extrude a stream that is being squashed by the nozzle tip which means that changes in direction will drag the center of the stream away from the center of the nozzle. this will cause different offsets from center when printing a straight line vs an arc. also as has been shown by the need for hardened nozzles to print any of the filled filaments due to rapid erosion of the nozzle tips there is a limit as to the precision you can define a nozzle and the life of a nozzle ( some of the high abrasive filaments can destroy a brass nozzle in hours ). standard practice is to print a single wall open top cube and measure the wall thickness with a micrometer ( cube not cylinder because measuring a flat surface is more accurate than an arc) and adjusting the extrusion multiplier to extrude exactly the width required. changing the stated diameter vs extrusion multiplier have exactly the same effect because what the slice uses for path determination is the programmed width and it uses the nozzle diameter and extrusion multiplier and layer height to derive the amount of material to extrude for each linear move.
RE: Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
@delamarre-frederic: No one wants to stop you doing something. It's still not clear to me how exactly you want achieve better accuracy. There are already tons of parameters in the slicer (I mean all major ones). Do you want to add a new parameter/calculation? Do you want to improve the hardware? Do you want to improve the firmware?
All PrusaSlicer (software) approximations are based on the hardware/firmware used. And currently that's the only way. Without feedback regulation you can't be certain how and how much came out of the nozzle. This is always a guess.
Often linked posts:
Going small with MMU2
Real Multi Material
My prints on Instagram
RE: Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
PLA is VERY pressure dependent: when it melts it is a FLUID and becomes susceptable to hydrodynamic forces..
I'm not talking about the fluid itself, just about parameters we can play with.
Linear advance correct the pressure in the melting zone to anticipate flow variations.
Rigid filaments like PLA are less dependant of linear advance than flexible materials, that's just what i said.
By printing single wall cylinders and study the result quality, you can precisely adjust the ideal width of extrusion.
Which is why I said measuring a flat is much preferable to measuring a cylinder to maintain experimental accuracy.
I never said i want to measure this cylinder, i'm just talking about visual inspection of the surface finish.
Next, by printing plain cyliders or rings with at least 8 outside and inside perimeters and study the surface finish after some layers ( to let the machine compensate the first layer potential over extrusion), i can look for my mirror surface finish by tweaking the nozzle diameter and correcting the flow by this method instead of tweaking it by the filament flow compensation.
You are ignoring the fact the bed is NOT stationary and can move underneath the nozzle. The frame and mechanics change dimension as they change temperature. It is unavoidable in the printer design. And unless you somehow map this displacement with extreme accuracy changes in layer dimensions can't be avoided and will nullify such testing/tuning. If the bed warps upwards 0.02 mm, your extrusion test is invalid (20% error with 0.1 layers). And the bed warps that much even with 60c temperatures; and much more when the bed is heated to 100c.
[Sarcasm on]
When i read this, the only thing i can conclude is : 3d printers are fucked up by construction, why am i boring with this machine ?
Many thanks tim-m30 to light my mind.
[Sarcasm off]
More seriously, is nozzle diameter involved in flow calculation or not ?
RE: Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
If nozzle diameter is not involved in flow calculations, my theory is dumb and i assume that.
If nozzle diameter is involved in flow calculations, it needs to be tweaked to improve accuracy.
That's the point.
About the slicer, i' m wondering why flow compensation or extrusion multiplier is filament dependant and not printer dependant.
It's the same question than retraction that is printer dependant, and finally Prusa slicer 2.1 has a retraction setting which is filament dependant (if i don't make mistake about my readings on github).
RE: Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
If nozzle diameter is not involved in flow calculations, my theory is dumb and i assume that.
If nozzle diameter is involved in flow calculations, it needs to be tweaked to improve accuracy.
Looks like you're talking about PrusaSlicer only. You set the nozzle diameter under the printer settings tab and adjust the extrusion width for different things under the print settings tab. For easy overall adjustments you're using extrusion multiplier under the filament settings tab. Disregard the position where it's located, it's just historically based. So you have a lot of parameters for the flow calculation based on the nozzle diameter.
There are NO parameters for nozzle shape. This is the most interesting part where the fluid dynamics would come into play. As far as I remember we always assume a linear extrusion (push 0.5mm, get 0.5mm extruded more or less immediate).
So my question again: What exactly do you miss or would like to make different/new?
Often linked posts:
Going small with MMU2
Real Multi Material
My prints on Instagram
RE: Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
Ok nikolai-r, i think i missed something :
amount of plastic extruded is only final shape dependant. You want a rounded edge rectangle with a precise section at a certain speed and slicer calculate the volume to be extruded.
There's no fluid dinamyc involved in our slicer calculations, it's not nozzle shape dependant, it's only a linear function.
So my theory about tweaking the nozzle diameter is dumb.
I have checked this as i'm printing samples with very different nozzle diameter and it has no effect.
I was certainly hallucinating by too many hours in front of my printer.
RE: Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
nozzle diameter affects 3 real life things, minimum reasonable line width ( lines smaller than nozzle diameter are not very controllable), max particle size of filament ingredients ( fibers/glitter/flakes), and nozzle tip exterior diameter sets max reasonable line width ( wider lines will not be height constrained by the tip outside of the bore)
RE: Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
nozzle diameter affects 3 real life things, minimum reasonable line width ( lines smaller than nozzle diameter are not very controllable), max particle size of filament ingredients ( fibers/glitter/flakes), and nozzle tip exterior diameter sets max reasonable line width ( wider lines will not be height constrained by the tip outside of the bore)
I'm aware of that...
Sometimes, when you're facing a problem with multiple parameters, you lost your basics.
I was in this situation when i developp my dumb theory.
RE: Flow, real nozzle diameter, dimensionnal accuracy and slicer strategy considerations.
More seriously, is nozzle diameter involved in flow calculation or not ?
The answer to that question is actually "not really" ... the flow calculation is the desired trace width times trace height time trace length, VOLUME, and the speed at which the trace is extruded is VOLUME/s. The nozzle only directs where the flow goes.