Extruder 102: Measure your hotend performance!
ฝัง
- เผยแพร่เมื่อ 24 ก.ย. 2021
- Let's add some experimental data to our analytical extruder stepper model! Thanks to henrikssn for the idea to use stepper torque data to calculate the values.
Link to Hotend Characterization Spreadsheet
github.com/eddietheengineer/d...
Follow me on Twitter: / eddietheengr - วิทยาศาสตร์และเทคโนโลยี
Hello everyone! I made a typo at 8:25, the Filament Force should be Torque * Gear Ratio / Radius, not multiplied by Radius as it states. The plots at the end are still correct though!
Hey Eddie, thank you so much for this content. Seriously. I hope that as a community we can encourage you to discover and document more and more data like this 👌
Loving what you guys do
Cool video, looking forward for the next one! One thought on max flow: as the extruder is close to skipping, we likely lose all resolution provided by microstepping. That's OK for infill and internal perimeters, whereas for the external one you could calculate a flowrate that required only about 10-20% of the motor torque. Similarly for the XY axes motors during acceleration/deceleration. Depending on extrusion, either the extruder or the XY motors could be the capping factor.
On a separate note, pushing the extrusion to the limit would likely reduce the actual extrusion.
I understood very little, but I know you're saying something big brain, so I hope you too can connect and collab!
If you're using microstepping for resolution instead of for smoothness of motion then you're doing it wrong in the first place. That introduces more cumulative loss of accuracy than it grants improved precision. If you need finer precision, get a motor capable of finer full steps or use more precise mechanical parts like smaller pulleys or a screw with a shorter pitch/lead, or gear it down. This is equally applicable whether you're pushing and pulling a carriage along a linear axis or feeding material with a rotary one.
Your content is straight out great! I really hope you'll get back to producing more content in the future!
Really interesting video as usual Eddie! Watching this made me realize that I haven’t actually stationary flow tested my magnum+ and worm drive extruder and need to do so.
Love it Eddie! Keep up the great work!
Great video Eddie! I think this shorter, "one idea, tested" video style works really well for you.
Stuff like this reminds me how nice it would be to have more data while printing (say, from a strain-gauge extrusion force sensor), not just to automate the generation of graphs like yours, but to usefully debug.
Thanks for the comment! It’s hard to know where the happy middle ground is a lot of times 😄
Excellent video. I'd be very interested in you experimenting with a volcano hot-end block. In my tests, I get higher flow rates than the Dragon non-HF hot-end.
Great video Eddie. In the past I used the Sabic MG94 ABS with great success to overcome flow rate issues. Its viscosity is way higher than other grades of ABS and you can print it at 280 degrees being dry with no issues increasing the layer adhesion. It would be an interesting comparison to make with your current material.
I love these videos keep it up 👍
Great video. For me personally, I found it easier the interprete the graph when the two plot axis being swapped. Thus, giving flow rate as a function of motor torque.
However, this is a minor change and does not change the given values.
I understood 80% of that.. good work
Thanks!
I think you could use the stallguard values reported by the extruder TMC driver to fairly easily estimate the torque required.
Nice video, thank you :)
Thanks for the education, I understand most of it from my college days (25 years ago), but I could not see the correlation of R^2=0.9764. On the extruder force vs flow rate. (is it stats on line of best fit?).
Wow Eddie, great video! Would there be any value in weighing the extruded filament after each run? Then you would have a quantitative result rather than a qualitative result, right?
Definitely--this is a very "brute force" way of testing things and has a lot of pretty major assumptions that can drive inaccuracies. There are a lot of ways to make it better!
@@eddietheengineer I think CNC Kitchen did a video to demonstrate max flow rate of a v6 hotend, and he simply weighed the extruded material at each flow rate. That is what convinced me to upgrade from the good old v6.
hi eddie i'm new to klipper fw i want to ask if i change the nozzle 0.4 with a bigger nozzle 0.6 or 0.8 do i need to change the nozzle diameter on the printer.cfg or just the slicer setting?
This is a great question! The nozzle diameter in the config to the best of my knowledge only affects the maximum extrude amounts. I'd recommend updating it especially if you're switching to a larger nozzle so that you don't get any errors.
One thing though is that if you've tuned PA with a smaller nozzle, you'll want to retune PA for the larger nozzle--your PA will not need to be as high!
Hope that helps!
Great video. Sometimes the stepper motor can skip a step without noticing. Did you consider using a closed loop stepper driver to detect skipped steps?
It would be even possible to automate the whole process (changing current, detecting skipped tests, increasing speed, ...) and test different hotends.
Yes! It would be fantastic to automate everything, some people have shown interesting in trying to make a script for this
Nice work! Increasing radius usually reduces force from torque..
Yes! That's why the extruders like Orbiter that use the 8mm Bondtech gears (34mm rotation distance) have a higher gear ratio (7.5) than the BMG with 5mm Bondtech gears (22mm rotation distance) with a 2.94 gear ratio
@@eddietheengineer shouldnt it be F = torque/radius? Cause the longer the moment arm the smaller the force.
Yes!! you're right, now I see where the typo is, thanks to you and Alan, I didn't catch it at first. Fortunately my excel spreadsheet is still correct, just my slide is incorrect. I'll see if I can overlay something at that point in the video.
Any reason you did not use design of experiments. Given it is an X^2 relationship it should give a pretty good curve fit.
There's a few things here that can make the curve fit not great. First of all--the TMC driver doesn't set the current to the value you specify, it sets it to the nearest of 32 steps that it has available. I'm getting those numbers together now, but unfortunately the default sense resistor for TMC2209 drivers doesn't really give much resolution in the low current range. Second of all, I'm testing in pretty coarse steps, 1mm/s increments don't give a ton of resolution but it's close enough for what I need. Third, there are a lot of assumptions about how linear torque is with stepper current, how significant the drivetrain losses are, how accurate the torque values are in general, how much the filament is slipping--the list goes on and on! The goal here isn't a perfect correlation, but more a semi-accurate model that we can use to show general direction.
@@eddietheengineer I am arm chair engineering here rather than firing up my 3D printers and agree it is just going to give a correlation. I would think a standard direct drive (no second gear) would be best to reduce measurement error.
With DOE it was more you mentioned alot of variables and that is the solution. It would be more helpful in controlling experimental parameters and looking for noise in the results which might be useful for modeling as much as anything.
Definitely! If I had the time and was motivated to do a DOE, that would be best. If I went that far though, I'd likely end up measuring force somehow instead of back calculating the force from an assumed stepper torque. In the end I'm fine with some pretty large error ranges since I'm doing this for fun and not a formal research project!
Have you tried to see if a exponential function fits you data points better?
I believe the pressure drop through the nozzle follows flow^2 per Bernoulli’s equation which is why I mainly picked that fit, but I’ll admit I haven’t done the math for that specifically
@@eddietheengineer Bernoulli only applies to incompressible fluids, although it could be a reasonabke approximation I think a different function could explain the data points better.
Generally speaking in the literature the viscosity of a polymer melt is classified as shear thinning. Personally I have observed that TPU melts behave shear thickening, however that obseveration is related to 3D printing, which means nozzle geometry could play a significant influence.
Regardless that is just nitpicking on my part. I really enjoyed your investigation.
@@DropEffect thanks for the context! You’ve definitely done more research into the hotend physics than I’ve done, my main goal here is to create a very simplified model that captures the general jist of what’s going on without diving too deep into the weeds. Once that’s done, maybe I’ll go down another rabbit hole 😄
I was wondering what about Shear thinning as well when I saw the trend line. If we look at the drag equation it boils down to two variables that we care about - velocity squared and viscosity. I wonder if our extrusion force is too low for shear thinning to dominate the increase in drag due to the increase in velocity.
However, it's definitely possible for shear thinning to dominate for a 3dp. I was just talking to Erik Gjovik who worked on the essentium HSE180 and he mentioned briefly that they are able to utilize on Shear thinning with standard nozzle sizes.
A side note - the drag equation is essentially bernoulli's but it is able to account for the compressibility of a fluid so I think Eddie's original reason for selecting the trend line makes sense.
Don't we also control the temperature?
CNCkitchen looks now like kindergarten to me.
Your force equation is incorrect. You have Force = Torque * GearRatio * Radius. It should be Force = Torque * GearRatio / Radius.
A quick way to show this is to check the units, your force will have units Nm^2.
Check the pinned comment 😉