Maximum extrusion rates and systematic under extrusion

[illuminarti 18]

As I have recently mentioned over on the Google Group, I've been doing some testing of extrusion rates (measured in terms of cubic mm of plastic per second), and how they are affected by print temperature. In particular I wanted to find the maximum sustainable extrusion rates for my Ultimaker.

I was able to find these limits, but I also noticed something else interesting - that at virtually all extrusion rates, the v2 extruder drive seems to exhibit under-extrusion of as much as 10 or 20% - amounts that gradually increase as the extrusion rate goes up.

This seems to be due to back pressure in the print head causing the filament to slip backwards during extrusion. Even though the filament isn't getting damaged in the ways that I normally associate with head blockages and extrusion problems, the spacing of the teeth marks on the filament becomes smaller at higher extrusion rates.

Effectively, it seems that the necessary steps-per-e parameter changes as we print faster. You can find the full details here:

http://www.extrudable.me/2013/03/29/exploring-extrusion-variability-and-limits/

 

[jeremy 0]

I'm glad someone has done some work on this! I've noticed a lot of underextrusion as well - moreso with some material than others.

Some material obviously has higher viscosity at a given temperature than other material. Also, though, it seems that the hobbed bolt can apply more force to some filament - I'd guess that this is a function of the material hardness.

It would be interesting to see what could be done with a more elaborate filament drive mechanism... say with 2 or more larger diameter wheels simultaneously pushing the filament.

Do you mind saying where you got the material from and what grade/name it was sold as?

-Jeremy

 

[karabas 1]

Try better extruder - with bigger roller diameters. F.e, I made similar tests on my seemecnc and I got skipping steps, I can hear it easily.

 

[illuminarti 18]

I agree that the exact results are going to vary with different filaments - both because of their fluid characteristics when molten (depending on colorants, additives, etc), and also due to their mechanical characteristics at the drive end. There are probably two different-but-related effects that need to be considered - both the ability to actively drive the plastic forward, and the ability to stop it slipping backwards (i.e., push harder, and hold on better).

A better extruder design would certainly help with the under-extrusion, and probably alter the maximum throughput rates too, but it doesn't alter the fact that most users have the current design, and are probably going to experience problems with under-extrusion. Nor does it alter the fact that there is an active limit to the amount of plastic that can be extruded per second, and we need to pay attention to that when selecting slicing parameters.

The plastic I was testing with is a sample batch that is not yet commercially available.

 

[joergen 2]

Thanks so much for the analysis. I have a couple of questions, comments, and wanted to see what your thoughts are:

1. you used a 0.65mm nozzle (0.332mm^2), which is a rather unusually large nozzle for the average UM user (0.4mm or 0.126mm^2), and even larger than my 0.56mm makergear nozzle (0.246mm^2). for simplify sake, one could say the standard UM nozzle is 1x, the makergear nozzle is 2x, and your nozzle is 3x larger. Since your test is exploring the max through put (including the max melting capacity of the 40W heater), I would say using a 3x nozzle is difficult to compare in this context?

2. our previous assumption was that a 0.4mm nozzle has a max capacity of about 10mm^3/sec, yet your 3x nozzle can't even reach that limit?

3. PLA behaves funny (I am saying this as a almost exclusive ABS user)

4. during my last plug event with a V2 hot end before I went back to my own custom V1 hot end, I found the V2 bolt (September 2011) be more effective transporting than the V3 bolt, at least with ABS... I didn't check the extrusion rate with PLA.

Another method to check the extrusion rate is solid infill, and careful observation:printing a flat object (i.e. 5x5cm), the solid infill should normalize after 2-3 layers at a certain speed. and the user can increase or decrease the flow until the infill is perfect, no ridges upwards (sign of over-extrusion) or gaps (sign of under-extrusion)... the flow rate (in %) gives a straight number how much to adjust.

The downside of this is that a uniform flow adjust only works with fixed speed slicers, such as cura/SF. for all other slicers, where the speed can vary greatly, the flow rate at a given speed would need it's own factor.

I think we had this question 1.5 years ago, but nobody had a conclusive answer, so thank you for for laying the scientific groundwork.

I think one could derive a "melt factor", which could be specific to PLA, plus a heat factor, since I think we can see the limits of the 40W UM heater in your chart already. I use a 50W heater, and seem to have more headroom than you have (not based on actual data, but more a gut feeling and experience), at least with ABS.

good work, can you rinse&repeat with 0.4mm? ;-)

 

[illuminarti 18]

Joergen -

Many thanks for the detailed feedback. You make some good points, and I'm certainly keen to learn from someone who has been exploring this a lot longer than I have. Here are my thoughts on your points:

 

1. you used a 0.65mm nozzle (0.332mm^2), which is a rather unusually large nozzle for the average UM user (0.4mm or 0.126mm^2), and even larger than my 0.56mm makergear nozzle (0.246mm^2). for simplify sake, one could say the standard UM nozzle is 1x, the makergear nozzle is 2x, and your nozzle is 3x larger. Since your test is exploring the max through put (including the max melting capacity of the 40W heater), I would say using a 3x nozzle is difficult to compare in this context?

 

Yes, its a much larger nozzle (I drilled it out myself with a 0.6mm drill bit, but best I can tell, I ended up with a slightly larger hole that I estimate to be 0.65mm.). So the results are definitely not directly applicable to a smaller nozzle, and you are right that there may be other factors coming in to play. There might be issues related to how quickly the plastic can absorb heat and melt when entering the head so fast, but in general, the temperature of the block seemed to stay constant within a degree or two during the tests, so it wasn't a problem of insufficient heater capacity overall.

 

2. our previous assumption was that a 0.4mm nozzle has a max capacity of about 10mm^3/sec, yet your 3x nozzle can't even reach that limit?

 

The upper limit that I saw of 24.3 would equate to a limit of 9.2mm³/s for a standard nozzle (based solely on opening size). And I have worked with a couple of folks who were having chronic extrusion failure problems when running at exactly that speed on a standard nozzle, and which improved dramatically with reduced volume. So, I think that that results may be more generalizable to smaller nozzles than you might think, when scaled appropriately. (I was also able to get slightly over a 10mm³/s throughput at 240º and with a little higher extruder spring tension - but at that temp the filament was too runny to get reliable measurements, so I gave up on those tests.)

 

3. PLA behaves funny (I am saying this as a almost exclusive ABS user)

 

Let's have none of those sort of prejudices here, thank you :smile: (But yeah, it does, kinda).

 

4. during my last plug event with a V2 hot end before I went back to my own custom V1 hot end, I found the V2 bolt (September 2011) be more effective transporting than the V3 bolt, at least with ABS... I didn't check the extrusion rate with PLA.

 

Yes, I'm just doing my tests with the v3 bolt. It certainly seems to behave oddly. I was surprised that the filament could have clean teeth marks, and be largely undamaged, and yet still be slipping quite so badly. It seems that the teeth are just too far apart, such that there are intermediary positions in which the filament can slip back past them? The extruder gear still turns exactly the right distance - the motor isn't slipping, or being pushed backwards as far as I can tell - the filament is simply able to move backwards past it at sufficient head pressure.

 

Another method to check the extrusion rate is solid infill, and careful observation:printing a flat object (i.e. 5x5cm), the solid infill should normalize after 2-3 layers at a certain speed. and the user can increase or decrease the flow until the infill is perfect, no ridges upwards (sign of over-extrusion) or gaps (sign of under-extrusion)... the flow rate (in %) gives a straight number how much to adjust.

 

I tried to make the test as objective and repeatable as possible... but to be honest, reliably measuring the weight of less than 1g of plastic was a painful and rather hit-or-miss affair that introduced its own margin of error. It's not going to work well for the lower volumes associated with the smaller nozzle, unless I run the tests for even longer. And it already takes far too long :smile:

Now that I better understand the mechanics of it, and the fact that the input process seems to degrade very gracefully with speed (up to the point where it totally fails, at least), I think that in future it will be sufficient to just measure the input length of filament at each rate and temperature. This can be done more accurately, and faster, than trying to capture and weigh a small piece of extrudate.

 

The downside of this is that a uniform flow adjust only works with fixed speed slicers, such as cura/SF. for all other slicers, where the speed can vary greatly, the flow rate at a given speed would need it's own factor.

I think we had this question 1.5 years ago, but nobody had a conclusive answer, so thank you for for laying the scientific groundwork.

I think one could derive a "melt factor", which could be specific to PLA, plus a heat factor, since I think we can see the limits of the 40W UM heater in your chart already. I use a 50W heater, and seem to have more headroom than you have (not based on actual data, but more a gut feeling and experience), at least with ABS.

 

Yes... I think it should be possible to at least approximate any speed-related under-extrusion factor in the slicer or firmware. Although, making it respond more dynamically to changing speeds would be trickier, given that it does seem to take some time to stabilize, and the Bowden tube introduces additional variability in the filament path, and maybe a little compression of the filament itself over the large length, that might be hard to deal with. We would probably need to find ways to test response to speed change over time... your solid infill test might be a good way to begin exploring that - seeing how quickly the extrusion quality responds to changed print speeds... hmmmm... that's a test for another day.

 

good work, can you rinse&repeat with 0.4mm? :wink:

 

Yes, I think that may be this weekend's project :smile:

 

[illuminarti 18]

Another thought on the comparison of large and standard nozzle throughputs... and the generalization of these findings to a smaller nozzle... depending on how the '10mm³/s' number was arrived at they could still be in strong agreement. Because in the case where I am hitting my 24mm³/s maximum throughput, I'm actually asking for 30mm³/s. That's the equivalent of 11.4mm/s on a smaller nozzle... so in theory, with this particular plastic at 230º, I should be able to print reliably at speeds of 30mm³/s - albeit I'm being saved from hitting the 24mm³/s limit, by the fact that the printer is chronically under-extruding by 20%.

So similarly (according to my tests) you could ask for 11.2mm³/s from a standard nozzle, and still get a complete (if slightly under-extruded) print without the filament grinding, or other obviously bad things happening.

Perhaps for practical purposes the limits should be expressed in terms of the requested print rate, rather than the attained one - although that seems to rely too much on the vagaries of the extruder drive - although maybe that is the main factor in this after all. Its not clear to me whether the limiting factor is to do with the shape of the nozzle and the properties of the molten plastic, or merely the amount of pressure the extruder drive can deliver. Would a better drive mechanism be able to deliver significantly higher throughput, even with the same hot end, I wonder. My gut tells me it would.

 

[joergen 2]

I spoke briefly with Jonathan, and he said this phenomenon is not limited to the Ultimaker, other printers experience this as well, in a very similar fashion, including the "direct drive" family of printers (sans bowden) which by default can exert a much greater extrusion force... which leads me to believe that it is more related to the pressure and melt capacity of the hot end.

The melt capacity is not the temperature fluctuation during extrusion. I don't think you will ever see the energy drop caused by the moving filament represented in the TC readings, at least not in the volume we are dealing with.

But plastic still needs a certain amount of joule/mm^3, and the energy necessary to do this is not uniformly applied, but somehow "radiates" from the surface of the filament to the core of the filament, when inside the melt chamber.

I think the heat reservoir of the nozzle plays a bigger role here as well, and I would venture that the larger mass of the MG nozzle helps in this context (6gr vs 4gr for the UM nozzle).

If you look at the J-head, or the version Taylor Alexander designed last year, they all have in common to "encase" the nozzle into the heater block, or actually making the nozzle part of the heater block, to increase the melt capacity where it matters most, the nozzle tip, and eliminating the thermal gap between the alu and the brass.

The alternative is using thinner filament: 1.75mm has a larger surface to volume ratio, and reaches a homogenous temperature faster than 3mm filament, and therefore requires a smaller heat reservoir in the nozzle.

so taking your experimental results, and combine them with my hypothesis above, I think faster extrusion on the UM needs a larger mass in the nozzle, and a better heat transfer between the alu and the brass (in conjunction with limiting the flow of heat upwards, to not prematurely melt the plastic where it shouldn't.

If you repeat your test with the 0.4mm nozzle, I would guess you will see a significant improvement when you apply some heat sink component between the nozzle and the alu.

 

[joergen 2]

On the other hand, it is easy to spot the phenomenon out there in the wild, with other printers, just by looking at the (shitty) infill like this one: http://www.instructables.com/id/How-to-Make-a-3D-Business-Card-With-a-3D-Printer/step5/Congratulations/

 

[snowygrouch 1]

This is a great thread, my opinion would be that if you apply back pressure to any drive system you

will inevitably get reduced extrusion. This is because of the highly elastic nature of the material

being worked on. I think that in the short term, a big improvement can be to have two driven

knurled wheels, rather than one.

However I think that really great results can only be achieved by having a 3rd wheel with a rotary pot (or something) just "downstream" of the knurled drive wheel, that will measure the linear speed of the filament passing through.

This would also really need to be coupled to a sensor detecting filament diameter in real time, to get really perfect

extrusion at all times.

I would be happy to help to do some mechanical design on this stuff, but I have no idea how it could be

integrated into the functioning of the firmware/arduino etc.....

C.

 

 

[owen 19]

Great work Simon.

The graph at the end says it all.

Thanks for your effort.

 

[theblur 0]

Hi everyone! not sure if this is the right place to ask this question or if the answer is somewhere in the forums already, if so I'd greatly appreciate a point in the right direction.

I've been printing now for about a year, fixed lots of under extrusion problems due to the filament wheel, tensions and blocked nozzles etc. However I have noticed a consistent line/layer of under extrusion in every one of my prints at around 1cm from the build plate. (after this point the print is fine)

The only thing I can think of thats causing it is dirt or dust or something stuck on part of the plate mechanism. I've cleaned it, re-oiled it but still get the extrusion line!? Any ideas?